CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES

diligentdeputyManagement

Nov 8, 2013 (3 years and 9 months ago)

85 views

CORRUPTION AND CONFLICT
IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
Horatiu A.Rus
¤
University of British Columbia
(This version:December 2007)
Abstract.The documented link between natural resources and civil con°ict is not well understood.
This paper uses a political economy framework to explore the emergence of resource-based civil
con°ict driven by group-level discontent.Previous models of resource con°icts are premised on the
idea that the desire of enrichment by appropriating resources is the driving force for insurgents.
This approach,however,treats the management of the contentious resources as exogenous,while
also failing to account for the grassroots dissatisfaction that is often reported to spark rebellions.
The proposed theoretical model o®ers a policy-based alternative:under certain conditions related to
the quality of governance,discontent about resource management can be instrumental in increasing
the likelihood of an insurgency.While in°uential contributions in the literature tell the`resource
abundance implies opportunity,implies greed-based con°ict'story,this paper focuses on relative
scarcity to justify discontent and prompt a`grievance-based'rebellion.The resource policy arises
endogenously as the corrupt government trades o® industry contributions and the cost induced by
manifestations of resource-related discontent.Conservation e®ects of both internal pressure,in the
form of civil unrest,and external pressure in the form of international trade and aid measures are
analyzed in turn,and regulator corruption is shown to be an important ingredient of con°ict.The
theoretical hypotheses put forth are tested on a comprehensive panel dataset in the empirical section
of the paper.In particular,resource depletion,the quality of governance and their interaction are
found to be signi¯cant and robust determinants of civil con°ict incidence.
JEL:Q27,Q56,D74,H56
Keywords:resource con°ict,quality of governance,political economy,renewable resource man-
agement,environment,international trade.
I am especially indebted to my supervisor Brian Copeland,as well as to my co-advisors Sumeet Gulati and Werner Antweiler for
invaluable support and guidance throughout this process.Many thanks for their encouragement and useful suggestions to Michael
Vlassopoulos,Patrick Francois,Matilde Bombardini and Hugh Neary.I also thank Per Fredriksson and the participants to the 2007
European Association of Environmental and Resource Economists (EAERE) Annual Conference in Thessaloniki,Greece,the 2007
Canadian Economic Association Conference in Halifax and the Development,Institutions,Environment and Trade (DIET) seminar
participants at UBC.The usual caveat applies.
¤
PhD Candidate,Department of Economics,University of British Columbia,997-1873 East Mall,Vancouver,BC,Canada V6T 1Z1.
Email address:horatiu@interchange.ubc.ca,webpage:http://grad.econ.ubc.ca/horatiu.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 1
1.Introduction
Economic policy-making is often responsible for the disenfranchisement of some groups in the society.
There is usually more than one interest group vieing for the regulator's favours,and systematically dis-
criminating against one side has the potential to generate con°ict.Economic discontent may erupt into
violence where the obstacle to a peaceful resolution of disputes is di±cult to circumvent.Many economic
activities can simply re-locate when the business environment is not satisfactory,so that tensions due to
regulation or re-distribution do not accumulate.Natural resources are,however,highly location-speci¯c,
and thus grievances are more prone to escalate into open con°icts.There are numerous accounts of the
link between natural resources and the prevalence of violence.`[In] many cases,central governments
promote unsustainable mining,logging,ranching,and other projects...[which] typically help prop up
unrepresentative,sometimes repressive regimes and enrich national elites and foreign corporations,with
few bene¯ts accruing to those whose lands are devastated.'
1
Natural resource management decisions,in
both the developing and developed countries,are still signi¯cantly skewed towards meeting political goals.
Resource-related violence is an important phenomenon.According to some estimates,more than one
in ¯ve con°icts worldwide are resource-based,
2
and civil con°icts dwarf inter-state con°icts in terms of
casualties,duration and number of participants.
3
Moreover,there is substantial case-study evidence of
resource con°icts triggered by an insensitivity to local concerns of an alliance formed by an opportunistic
government and the extractive industry.
4
This paper uses a political economy framework to explore how con°ict can occur if resource management
policies fail to address particular interest group concerns.Although the most common narrative of the
origins of most civil violence episodes involves some form of group grievances,existing economic models
of con°icts are instead premised on the idea that the desire of enrichment by resource appropriation is the
cause of most rebellions.Moreover,this approach fails to account for the critical support for insurgency
of a signi¯cant share of the population that is often apparent in the initial phases,
5
while the empirical
support is not robust.The theoretical model proposed here o®ers a policy-based alternative:legitimate
grievances related to the way in which resources are managed can translate into violence.While the
appropriation mechanismlacks a clear policy recommendation,the policy channel prescribes that devising
1
See Renner (1996),p.59.
2
See Renner (2002),p.6.
3
See Fearon and Laitin (2003),p.75.
4
E.g.in Indonesia,the Aceh con°ict against the government and ExxonMobil,in Papua New Guinea,the Bougainville con°ict
against the government and the mining corporation RTZ,in Nigeria,the Niger Delta con°ict against Royal Dutch/Shell and other
Western oil giants and the government are just the largest ones.See Regan in Ballentine and Sherman (eds) (2003),p.133-166 and
Renner (2002),p.40-47.
5
Often,declared motives may also proxy for additional underlying causes,while in other instances claims for socio-political justice
may mask predatory intentions.E.g.the separatist Papuan Freedom Organization`did not gain much support from the local population
until the 1970s,when it harnessed grievances against a large-scale mining operation.'Renner (2002),p.43.
2 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
a more inclusive resource policy has the potential to reduce the incidence of civil con°icts.The focus is ¯rst
on the government's resource policy,a function of the quality of governance.The resulting environmental
outcome may spark legitimate resource-related discontent,regardless of the eventual unfolding of the
con°ict or shifts in rebels'motivations.Resource depletion may also have regional consequences,for
which international intervention might be appropriate.Hence,the model also investigates the e®ects of
international trade measures and of international transfers on such an environment.Speci¯cally,the paper
asks whether these`external pressures',which may complement the`internal'constraint represented by
the possibility of domestic violence against a corrupt government,are instrumental in curbing depletion,
thus indirectly reducing the likelihood of con°ict.
There are three agents in this economy:the self-interested government which is responsible for mak-
ing resource management decisions,the large scale industry which in°uences policy through regulation-
contingent payments,and the small-scale subsistence harvesters who live o® the resource and can migrate
to take payo®-improving outside opportunities.In a majority of such cases,the basic scenario is simi-
lar:wide-scale exploitation of a natural resource bene¯ts large corporations and the ¯nancially interested
government,while damaging the environment and threatening the way of life of local populations.
6
The main results of the paper are the following.Absent the shadow of con°ict,the opportunistic
government attaches a lower e®ective weight to locals'concerns and sets a larger harvesting quota than
optimal,and resource exploitation increases with the corruption of the government and with the rate of
time preference parameter.The specter of con°ict decreases exploitation in certain environments,where
parameters are such that the increased discontent of the locals dominates their induced emigration.While
corruption is generally detrimental to stability since it indirectly increases conservationist grievances,more
corruption could reduce the con°ict-generating e®ect of depletion,a somewhat counter-intuitive ¯nding.
Good outside opportunities available for subsistence harvesters fosters migration and has a negative e®ect,
while the level of population has a positive e®ect on the likelihood of con°ict.Provided the parameters
are such that the equilibrium level of the resource is not too low,international sanctions targeting the
resource management process in this environment where the government is corrupt and con°ict is possible
achieve their conservationist goal whenever they result in producer price reductions,while the e®ectiveness
of conditional international aid can be inversely related to the quality of governance.The quality of
governance plays an important role in assessing the e®ects of both external (trade or aid) and internal
(political violence) pressure.The empirical section uses a panel dataset including 120 countries over a
20 year period and analyzes the discontent-based civil con°ict hypothesis.While in the previous studies
6
`In many developing countries,the economic bene¯ts of mining and logging operations accrue to a small business and government
elite and to foreign investors.But in case after case,an array of burdens - ranging from expropriation of land,disruption of traditional
ways of life,environmental devastation,and social maladies - are shouldered by the local population.'Renner (2002),p.8.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 3
broad inequality and ethnic fragmentation variables were used to proxy for grievances,this channel is
more closely explored here by employing data on resource depletion.It is shown that income,population
level,the quality of governance,depletion and their interaction indeed appear as signi¯cant determinants
of civil con°icts and have the hypothesized signs.
Most studies on the relationship between resources and con°ict - and especially those proposing preda-
tory behaviour as the dominant mechanism for civil con°ict - look at easily lootable,high value non-
renewables,such as precious minerals.To o®er a complementary perspective,the present theoretical
model incorporates renewable resources,typically less prone to violent appropriation.
7
Most civil wars
occur in developing countries,where large groups depend on the environment for subsistence,
8
and it
is broadly agreed that`most life-giving [natural] resources are renewable'.
9
Alternatively,the model can
support the more general interpretation of being concerned with a population's discontent over the general
state of the natural environment as managed by a corrupt regime,including deforestation,severe overex-
ploitation of ¯sheries,soil degradation,water crises,and the two parties competing for the government's
favours can be called harvesters and conservationists.
10
The economics of con°ict is an emerging sub-¯eld in the discipline.Hirschleifer and Grossman pioneered
the theory of appropriative con°ict,where two symmetrical sides,choosing from the same set of actions,
optimally allocate their resources to production,soldiering and insurgence,and where con°ict arises as an
equilibrium outcome when parameters are such that a probability of insurgence success is high.
11
In this
tradition,Olsson (2007) presents a predator-prey model where the ruler is assumed to own all resources,
while the rebels'claimis predatory,and tests the theoretical predictions for the case of the rough diamond
industry.The study is complementary to the one in this paper,in that here the motivation lies in the
legitimate resource policy driven claims by marginalized interest groups dependent on the life-sustaining
resources.
Most models in this literature are essentially static,
12
while resource rents are exogenous.Prominent em-
pirical studies on the topic of resource-driven civil con°icts,starting with the series by Collier and Hoe²er
as part of the World Bank project on the Economics of Civil War,Crime and Violence,report economic
7
Renner (2002) notes that when resource wealth is the cause of con°ict,it is usually associated with lootable minerals,while scarcity
is more often linked with renewables (p.9.).
8
Civil wars in Indonesia,Cambodia,Burma,Liberia and the Democratic Republic of Congo have involved timber.In the Ivory
Coast,cacao and cotton,along with diamonds,are documented to have been connected to the con°ict.See the report by Global Witness
at http://www.globalwitness.org/pages/en/cote
divoire.htm.
9
See Reuveny and Maxwell (2001),p.720.
10
The natural processes by which the environment absorbs part of the industrial impact would correspond to stock depletion.This
broad view can accommodate situations where unrelated industrial or extractive activities damage the environment,like in the Niger
Delta,while the government fails to let locals share in the resource revenue bounty as compensation for their lost livelihood.
11
See Neary (1997) for a comparison between rent-seeking and con°ict models.
12
In a recent review,Gar¯nkel and Skaperdas (2006) point out that`only the surface of the dynamic e®ects of con°ict has been
scratched.'(p.54.)
4 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
opportunity - the`greed'motivation - as predominant explanation of internal violence.However,many of
these results are not robust to the de¯nition of con°ict,of resource-dependence or the particular sample
of countries or year of analysis.
13
The approach in this paper resembles the ones in Damania and Barbier
(2001) and Barbier et al.(2005),who present,respectively,political economy models of renewable and
non-renewable resource harvesting and focus on trade policy instruments and resource conversion.This
paper di®ers chie°y in structuring the interaction among three groups of players with distinct objectives
and in focusing on resource-based civil con°ict.The present model is inherently dynamic,featuring the
intertemporal management of a renewable resource (which could be broadly interpreted as the environ-
ment),endogenizes the resource policy and also allows for an asymmetrical treatment of the government
and potential rebel sides.
The rest of this work is structured as follows.The next section looks at the quality of governance as
driving policies susceptible to generate discontent and it presents the no-con°ict opportunistic government
equilibrium in contrast to the the`no-corruption'¯rst best benchmark.The following section allows for
internal con°ict to be a factor and studies properties of an equilibrium where costly manifestations of
discontent factor into the government's decision making.In addition,the section discusses how resource
depletion and the quality of governance may in°uence the emergence of full scale con°icts.The e®ects of
possibly complementary external conservationist pressure in the formof trade sanctions and aid conditional
on resource conservation performance are then analyzed.The third part of the paper tests the relationship
between the prevalence of civil resource-centered con°icts,the quality of governance and the management
of natural resources on a large panel dataset,while the last section summarizes the main ¯ndings and
concludes.
2.A Political Economy Model of Renewable Resource Exploitation
2.1.Peaceful resource (mis)management.
The three agents in the model are:the government deciding on the exploitation rate of a renewable
resource,the extractive industry acting like an organized group to lobby the government for higher than
optimal quotas,and the local peasants,each harvesting a subsistence amount of the resource.The paper
¯rst looks at the case where con°ict is not an option,possibly due to the fact that harvesters cannot
resolve their collective action problem.When subsequently allowing for the prospect of con°ict to play
a role,it is assumed - also without modeling that particular game - that harvesters have solved the free
riding obstacle and are able to challenge the government.
13
As Hegre and Sambanis (2006) contend in a recent study that undertakes a`global sensitivity analysis'of these studies p.509.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 5
The resource exploitation regime is modeled as follows:the industry presents the government with a
`bribe schedule'that relates contribution payments to harvesting quotas.
14
The government takes this
schedule as given and maximizes its own utility function that depends on the welfare of the sector and
the contributions it gets from the big industry to set the quota at a high level.
15
While the industry only
values the proceeds from harvesting,the locals'very existence is linked to the resource.
16
The way this
feature is modeled here is the following:the individual utility function of a local harvester is equal to
the amount of the resource required for subsistence
¹
h,less the disutility of e®ort required to harvest it.
Peasants are heterogeneous with respect to their harvesting ability q
i
and their harvesting function equals
the product of combining three inputs:the harvesting ability,the level of the resource stock and labour:
¹
h = q
i
Sl
i
.Their preferences are represented by a (disaggregated) utility function of the following form:
u
i
(
¹
h;l
i
) =
¹
h ¡l
i
=
¹
h ¡
¹
h
q
i
S
.
Peasants`migrate'
17
or pursue outside opportunities,represented by a unit labour wage w,
18
treated as
exogenous,if they can improve their payo®.In the present context this means working less in order to
earn the subsistence amount
¹
h.The individual peasant i remains a harvester as long as the utility she
would get by taking an alternative opportunity does not exceed the status quo payo®:u
i
(
¹
h;l
i
) ¸ u
i
(
¹
h;l
0
i
),
equivalent to l
i
· l
0
i
or
¹
h
q
i
S
·
¹
h
w
which implies
q
i
¸
w
S
,
and takes the superior outside option otherwise.As the resource gets depleted,the local resource-
dependent population decreases,and so does their group strength.
19
Assuming that harvesting ability is uniformly distributed in the total peasant population on an interval
de¯ned by two extreme values that are determined by an initial distribution and past migrations,q
i
»
U[q
;¹q],the fraction of the normalized initial population that remains at every level of the resource stock
is L(S;w) = (¹q ¡
w
S
),while the mass of city migrants equals (
w
S
¡q
).
20
The rural harvesting population
14
Similar,e.g.to Grossman and Helpman (1994).
15
The monetary contribution paid by the industrial sector to the government is called`bribe'.To be exact,however,this terminology
is only accurate from the point of view of the government,who embezzles the quota rights.From the ¯rm's point of view,`the bribe'is
just the cost of obtaining the harvesting licence.In theory,this should be illicit from the company's perspective when the amount paid
to the government would not be enough to compensate for the true social welfare cost of harvesting.However,no restrictive assumptions
are made in this respect here.
16
This responds in part to some critics of the economics models of con°ict contending that the exclusively extrinsic,payo®-motivated
players are incompatible with the many documented instances of intrinsically motivated actors.See e.g.Cramer (2002)`Homo Eco-
nomicus Goes to War'.
17
This does not have to involve physical migration,only disconnect from the resource.
18
The amount w can be thought of as the minimum wage in the neighbouring city,region or country where migration is feasible.
19
...although less than proportionately when it is assumed harvesting ability is positively correlated with ¯ghting ability.
20
Notice that the formula for the number of locals remaining is robust to subsequent migrations:as the low ability types leave,the
lower bound of the ability interval moves to the right,yet the upper bound remains ¯xed.
6 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
decreases as the resource is depleted and it becomes increasingly di±cult to provide sustenance to everyone,
and also decreases when outside opportunities become more attractive:L
S
> 0 and L
w
< 0.
Suppose the peasant population matters for the government,for instance in their role as voters,a
common assumption in the political economy literature.Then,even an opportunistic regulator`cares'
about social welfare,along with bribes.The aggregate utility of small harvesters is the sum of the utilities
of locals remaining in the rural area and the city migrants:u = u
r
+u
c
.To obtain the aggregate utilitarian
u
r
,integrate the individual utilities of the remaining locals on the relevant support as follows:
u
r
=
R
¹q
w
S
u
i
dq
i
=
R
¹q
w
S
(
¹
h ¡
¹
h
q
i
S
)dq
i
=
¹
h(¹q ¡
w
S
) ¡
¹
h
S
ln
¹qS
w
.Similarly,adding the welfare of migrants yields
u
c
=
R
w
S
q
(
¹
h ¡
¹
h
w
)dq
i
=
¹
h(1 ¡
1
w
)(
w
S
¡q
),so that the aggregate utility of peasants can be written as:
U(S) =
¹
h(¹q ¡q
) ¡
¹
h
S
ln
¹qS
w
¡
¹
h
w
(
w
S
¡q
).
It follows that the marginal utility of a resource unit is:U
S
=
¹
h
S
2
ln
¹qS
w
> 0,and the second derivative
U
SS
=
¹
h
S
3
(1 ¡2 ln
¹qS
w
) < 0.
The industrial and small scale harvesters always coexist in this model.The paper assumes for sim-
pli¯cation that the pro¯t function of the industrial sector is such that pro¯ts are positive or the ¯rm is
active for all relevant stock levels and prices.In other words,the extraction technology is assumed to be
`advanced enough'to yield positive pro¯ts even at low levels of the resource.For tractability it is assumed
that all agents discount the future at the same rate.The resource management instrument chosen by
the government is a quota on harvesting,which is the predominant form of resource and environment
protection policies worldwide.
21
We are interested in how the quality of governance a®ects the environmental outcome.As a benchmark,
consider ¯rst the ¯rst-best case where there is no corruption.Let ¼(p;H;S) be the industry's total pro¯t,
increasing in all arguments:the price of the ¯nal resource good,the harvesting quantity and the stock
of the resource,since harvesting costs are assumed to increase with resource scarcity.Denote by ± the
rate of time preference in the society,and by F(S) the natural growth function of the resource,which is
assumed to be strictly concave.An honest government maximizes the discounted present value of a social
welfare function of the following form:
Max
Q
Z
1
0
e
¡±t
[¼(p;H;S) +U(S)]dt subject to
_
S = F(S) ¡H ¡
¹
H and H · Q:
The ¯rst constraint is represented by the growth of the stock,which depends on a natural growth
function F(S) and the combined harvesting H and
¹
H,where the latter represents the subsistence amount
21
The quota does not bind the small-scale harvesters due to the impossibility of enforcement and/or the fact that the government
trusts the`environmentally responsible'locals not to over-deplete the resource,or the fact that the local subsistence harvesting is of a
much lower order of magnitude than the Total Allowable Catch (TAC) or,like in many communities in North-America,the existence of
treaties which allow harvesting privileges for locals.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 7
harvested on aggregate by the local peasants:
¹
H(S) =
¹
h(¹q ¡
w
S
).The second constraint provides that
industrial harvesting H does not exceed the total allowable catch Q and will be binding at optimum.I
assume lump-sum transfers are ruled out,perhaps because of a signi¯cant marginal cost of public funds,
and the only distributional channel is via the resource policy.
Optimal harvesting regulated by an honest government without the possibility of con°ict satis¯es the
following steady-state decision rule,as detailed in Appendix A(1):
(2.1) (F
S
¡
¹
H
S
) +
¼
S
+U
S
¼
Q
= ±:
This is a version of the augmented modi¯ed golden rule of renewable resource exploitation introduced
by Clark and Munro (1975).
22
The last term on the left-hand-side constitutes the marginal stock e®ect
(MSE).This arises since in this setting welfare depends not only on the °ow of harvesting,but also on the
stock of the resource via its e®ect on harvesting costs for the industry as well as for the locals.Drawing a
parallel to capital theory,the above equation states that it is optimal to invest in the resource by abstaining
from harvesting until the net marginal resource yield (F
S
¡
¹
H
S
),augmented to account for the e®ects on
industry pro¯ts and locals'utility,equals the social rate of discounting the future.In other words,the
left hand side represents the social return from conserving one unit of the resource,while the right hand
side is the social cost of doing so.Alternatively,the MSE can be re-written as the society's`marginal
rate of substitution'between keeping the resource in situ and harvesting it:if W
h
is the welfare function
of the honest government described above,W
h
Q
= ¼
Q
and the marginal welfare e®ects of harvesting and
resource level are calculated as:W
h
S
= ¼
S
+U
S
j
H=H
¤
,then the marginal stock e®ect MSE =
W
h
S
W
h
Q
.The
expression in (2.1) yields a unique stationary resource stock level achieved under an honest government,
which is denoted by S
h
.
23
When corruption is present,there is scope for strategic industry-regulator interaction.To keep things
standard,the menu auction introduced by Bernheim and Whinston (1986) is assumed as framework for
the game,as it has served as the workhorse model for most recent political economy studies,starting
with Grossman and Helpman (1994).The extractive interest group presents the government with a bribe
schedule that relates contributions to policy alternatives.The government then chooses the policy measure
by trading o® bribe revenues and losses in welfare brought about by the socially ine±cient choice induced
by the contributions.The ¯rms harvest according to the set quota level and pay the corresponding bribe
to the government.Denoting by G = W+¯B the utility function of the government,as a sumof the social
welfare and the value of the contributions,and by ¼ the gross industry pro¯t,Bernheim and Whinston
22
See also Munro (1979),p.5.
23
See Clark and Munro (1975),p.95.
8 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
(1986) showthat the`truthful'equilibriumharvesting has to satisfy the following:(C
1
):Q = argmax
H
fGg
and (C
2
):Q = argmax
H
fG+¼g.
24
If the above conditions were not satis¯ed,it would be possible for
the ¯rm to adjust its contribution schedule so as to induce the government to maximize the joint surplus,
while keeping most of the gain for itself.
25
An implicit assumption is that all agents are forward-looking,and that property rights are enforced.
26
This simpli¯cation allows one to avoid diverting the analysis towards free riding incentives and focus
on con°ict instead.
27
The local harvesters have no incentives for poaching,since they only take the
minimum subsistence amount,while the large extractive industry purchases its harvesting rights from
the government.Resource overexploitation by extractive companies is often sanctioned by successive
opportunistic o±cials,thus the briber becomes practically immune to expropriation due to the complicity
of the regulator.Under these conditions,it can be assumed that the agents are forward-looking and solve
an optimal control problem.
Condition (C
1
) above amounts to maximizing G(W;B) = W+¯B,where Wis the aggregate welfare,B
is the bribe and (¯ ¡1) a measure of corruption.The weight attached to the bribe entering government's
utility function is higher than unity ¯ > 1 when the government is corrupt,since it values a unit of income
in its hands more than when in the hands of the public.The government's problem is then the following:
(2.2)
Max
Q
Z
1
0
e
¡±t
[¼(p;S;Q) ¡B(Q) +¯B(Q) +U(S)]dt subject to
_
S = F(S) ¡Q¡
¹
H(S):
Since total ¯rm pro¯ts are increasing in the amount harvested H on the relevant range and the bribe
schedule is increasing in Q,the ¯rm will always have an incentive to lobby for a larger quota and the
implicit constraint H · Q holds with equality.The solution to the optimal control problem is presented
in Appendix A(2) and it yields a general decision rule of the following form:
(2.3) (± ¡F
S
+
¹
H
S
)[¼
Q
+(¯ ¡1)B
Q
] ¡¼
QQ
_
Q¡(¯ ¡1)B
QQ
_
Q¡¼
QS
_
S = ¼
S
+U
S
:
Similarly,according to condition (C
2
) above,the optimal harvesting quota and bribe tuple has to also
maximize f¼ +Gg:
(2.4)
Max
Q
Z
1
0
e
¡±t
[2¼(p;S;Q) ¡2B(Q) +¯B(Q) +U(S)]dt subject to
_
S = F(S) ¡Q¡
¹
H(S)
which is shown in Appendix A(3) to imply:
(2.5) (± ¡F
S
+
¹
H
S
)[2¼
Q
+(¯ ¡2)B
Q
] ¡2¼
QQ
_
Q¡(¯ ¡2)B
QQ
_
Q¡2¼
QS
_
S = 2¼
S
+U
S
:
24
See Lemma 2,op.cit.,p.10.A similar approach is adopted in Barbier et al.(2005).
25
See for instance Grossman and Helpman (1994),p.839.
26
Commercial exploitation of resources with a government-issued licence implies some system of property rights is in place.
27
See e.g.Hotte et al.(2000) for an endogenous property rights enforcement model.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 9
In a steady-state equilibrium,conditions (2.3) and (2.5) above become respectively equivalent to:(± ¡
F
S
+
¹
H
S
)[¼
Q
+ (¯ ¡ 1)B
Q
] = ¼
S
+ U
S
and (± ¡ F
S
+
¹
H
S
)[2¼
Q
+ (¯ ¡ 2)B
Q
] = 2¼
S
+ U
S
.Solving for
the optimal`bribing intensity'B
Q
=
1
¯¡1
(
¼
S
+U
S
±¡F
S
+
¹
H
S
¡¼
Q
) from the ¯rst and substituting into the second
yields the following truthful-contribution equilibrium decision rule in the steady state:
(2.6) (F
S
¡
¹
H
S
) +
¯¼
S
+U
S
¯¼
Q
= ±;
where recall that ¯ > 1 because the government is assumed to be corrupt.
28
This expression yields
implicitly the steady state stock of the resource under a corrupt government S
c
.The government allows
industrial harvesting to deplete the resource up to a level at which the discount rate just equals the net
marginal bene¯t of conservation to the government,with the industry pro¯ts being assigned a higher
weight than the locals'utility.Substituting the optimal exploitation rule in the bribe intensity expression
above yields:B
Q
= ¼
Q
¡
¼
S
±¡F
S
+
¹
H
S
,which is the adapted expression for local truthfulness:the gradient
of the bribe schedule coincides with the marginal contribution to the lobby's pro¯t of the extra unit of
harvesting,taking into account the stock externality of resource depletion.
29
The following intermediate
results are instructive for further analysis.
Lemma 1:Bribing intensity increases as the stock gets depleted.
While by applying (2.6) it is easy to see that B
Q
> 0,that is the bribe schedule increases in the quota,
the bribing intensity can also be expressed as:B
Q
=
U
S
¯
¢
1
±¡F
S
+
¹
H
S
.Then the following cross-partial
derivative of the bribe schedule B
QS
=
¯
(±¡F
S
+
¹
H
S
)
2
[U
SS
(± ¡F
S
+
¹
H
S
) +U
S
(F
SS
¡
¹
H
SS
)] is negative under
the assumption of negative second partial derivatives of the resource growth and pro¯t functions with
respect to stock.In order for the regulator to respond to its policy preferences,the industrial lobby needs
to keep it weakly better o® than under the ¯rst best case.When the resource is more scarce,the necessary
e®orts by the ¯rm to increase the harvesting quota by one unit are larger,due to the need to compensate
for a sharply falling social welfare function.
Lemma 2:A more myopic government is less conservationist.
This result is intuitive,and the proof provided in Appendix A(4)(i) is immediate.Higher discounting
of future periods implies larger weights attached to harvesting-derived present payo®s and consequently
leads to a lower steady-state resource stock level.
Lemma 3:If the marginal stock e®ect under one resource management regime is larger than under a
di®erent one at all possible stock levels,then the equilibrium stock level in the ¯rst case is also larger than
in the second,ceteris paribus.
28
The problem is equivalent in the steady state with one in which the government maximizes a social welfare function in which
industry pro¯ts enter with a weight of ¯,while the aggregate utility function of the locals enters with a unitary weight.
29
This is,for instance,the equivalent of condition (9),p.839 in Grossman and Helpman (1994).
10 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
In short:if 1 and 2 denote di®erent regulator objectives and MSE
2
(S) > MSE
1
(S);for all feasible levels
of stock (8S 2 [0;K]),then S
¤
2
> S
¤
1
.The proof by contradiction is provided in Appendix A(4)(ii) and
this result facilitates the comparisons of equilibrium stock levels across the di®erent policy environments.
Comparing the equilibrium stock level set by a self-interested government with the`e±cient'level set
by an honest government,i.e.equations (2.1) and (2.6) above,yields the following result:
Proposition 1:The steady-state equilibrium stock level when harvesting is regulated by a corrupt
government is always below that obtained under an honest government.
Proof:provided in Appendix A(5).¤
The proof relies on Lemma 3 and on the fact that the corruption coe±cient is larger than unity.This
proposition establishes the dynamic natural resource counterpart of the Lopez and Mitra (2000) result,
which states that pollution is higher in corrupt regimes,obtained in their case as a Nash bargaining
outcome of a game between the government and the private ¯rm.
30
This ¯nding is not as banal as it may seem,since an established result in the ¯eld of natural resource
economics is that pro¯t maximization and e±cient resource conservation are not incompatible under the
assumption that property rights are well-de¯ned.
31
The optimal decision equation for a corrupt government
can be re-written as:F
S
¡
¹
H
S
+
¼
S
+U
S
¼
Q
=
~
±,where
~
± = ± +
U
S
¼
Q
(1 ¡
1
¯
) is above ± for all values of the
corruption coe±cient larger than one.The presence of corruption is equivalent to the government being
more myopic in its resource policy making.Intuitively,the resource is valuable for a group to which the
corrupt government is attaching a lower weight.In the ¯rst best case when the regulator is honest,the
presence of small harvesters leads to an equilibrium stock that is larger than the pro¯t-maximizing level.
When the regulator is corrupt,the policy in°uence of locals is weakened and the second-best income
transfer that takes place via the resource policy entails a lower steady-state level of stock.It is evident in
the optimal decision rule formula that the opportunistic government e®ectively attaches a weight of ¯ > 1
to terms related to industry pro¯ts,while the harvesters'utility is assigned a unitary coe±cient.Hence,
the resource outcome is unambiguously less conservationist than under an honest government.
Thus far the behavior of resident small harvesters has been restricted to each consuming passively a
subsistence amount
¹
h of the resource good.Therefore,their channel of in°uencing the environmental
outcome was,by design,limited to the inclusion of their group utility in the social welfare function.
32
In
30
The result also corroborates the ¯ndings in Barbier et al (2005) for the case of resource conversion,and those in Fredriksson and
Svenson (2002) for environmental regulation.
31
`...the monopolist is the conservationist's friend.'according to Solow (1974),p.8,among others.
32
The above harvesters only count as voters,which is the reason why even a corrupt government takes their concerns into account
when devising the resource management policy,or alternatively they could`vote with their feet'by migrating.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 11
what follows,the harvesters will be in the position to exert additional in°uence on the government,as
they can generate unrest which is costly for the government.
2.2.Internal pressure:the shadow of con°ict.
Assume now that the locals have the option of manifesting their discontent by civil unrest or starting
an insurgency.
33
The trade-o® for the government is between allowing for more resource depletion in
exchange for monetary contributions from the industry and augmenting the discontent of the locals.
There are two features of the present view of con°ict that should be mentioned here.First,civil tensions
are modeled in the paper not as a discontinuity in the government's utility pro¯le,but rather as a process
that can be tolerated by a corrupt government as`a cost of doing business'.This allows one to look at the
interplay between the threat of con°ict and policy outcomes,when one interest group is not organized as
a lobby,but rather as`rebels'.The specter of con°ict becomes relevant not due to direct appropriation
of the resource,but rather via political channels,through the higher-than-optimal harvesting quota set
by the corrupt government.Unlike other papers in the literature
34
which assume perfectly myopic agents
that deplete the resource up to a con°ict-generating threshold,in this paper corruption is the actual
trigger of con°ict.Consequently,the issue becomes one of the degree of civil unrest,where all-out civil
war erupts when violence reaches a certain level,which may or may not be known by the government.
This is consistent with the empirical section,where con°ict is de¯ned in relation with a certain`battle-
related deaths'-threshold.This will be modeled here as an expected cost of con°ict exceeding an arbitrary
threshold.
A second distinguishing feature of con°ict as modeled here is that it is asymmetric.Most theoretical
models of con°ict feature two parties facing symmetric problems of allocating scarce resources - usually
labour,bounded by population sizes - between productive and bellicose uses.However,in episodes of
resource-based violence it is far more common to witness one group revolting against the government.Civil
con°icts are de¯ned in the leading global data collection projects as invariably involving the government.
35
The interaction between the two parties is then asymmetrical,with the government having non-military
options at its disposal to either alter the contentious resource policy in order to appease the potential
rebels,or to deter them by investing in the military and thus avoid an all-out con°ict.
The discontent of the locals vis-µa-vis the government-sanctioned exploitation of resources as modeled
here corresponds to`frustration-aggression'theories based on relative deprivation:violence arises as a
33
This situation can result from small harvesters resolving their collective action problem.Using the language of Acemoglu and
Robinson (2005),the small harvesters have de facto political power in this setting.
34
Maxwell and Reuveny (2000),Homer Dixon,Gledisch,among others.
35
UCDP de¯nes con°ict as:`a contested incompatibility that concerns government and/or territory where the use of armed
force between two parties,of which at least one is the government of a state,results in at least 25 battle-related deaths.'See
http:==www.prio.no=cwp=armedcon°ict=current=Codebook
¡
v4¡2006b.pdf for more details.
12 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
result of a discrepancy between actual payo® and perceived merit.
36
Small harvesters'maximum potential
payo® corresponding to the case when they rebel and win a con°ict against the government is what they
obtain when they actually set the harvesting policy themselves.It is reasonable to assume here they prefer
zero industrial harvesting,given that there are no employment and income spillovers from the industrial
harvester of the resource.The stylized fact that the local populations hardly ever draw any bene¯ts
from industrial resource extraction is con¯rmed by ample case-study evidence,and is mostly based on
widespread use of capital-intensive exploitation technology and of a relatively skilled migrant workforce.
37
Therefore,the ideal payo® of peasant harvesters is U
¤
= U(S
¤
),where S
¤
denotes the`pristine'state
of the resource where the only extraction is done by subsistence harvesters,or S
¤
solves F(S) =
¹
H(S).
38
The propensity to rebel ½ is de¯ned as a function of the stock-related objective discontent of the local
peasants,measured as ¢U = U
¤
¡ U.The likelihood of social unrest depends on the resource-related
motivation of the locals and it is represented by the following function de¯ned on the continuous domain
between zero and one:
½(S) =
U
¤
¡U
¢U +U
=
¢U
U
¤
:
Using the aggregate utility function de¯ned before,½ =
1
u
¤
[
¹
h(
1
S
¡
1
S
¤
) +
¹
h(
ln
¹qS
w
S
¡
ln
¹qS
¤
w
S
¤
)] and ½
S
=
¡1
u
¤
¹
h
S
2
ln
¹qS
w
.Consequently ½
S
= ¡
U
S
u
¤
< 0,so unrest is more likely as the resource gets more depleted.
Appeasing or deterring unrest is costly.Assume that the government's utility drop induced by the locals'
discontent is an increasing function of the size of the peasant population L:C(L(S;w)) = °L(S;w) =
°(¹q ¡
w
S
),with ° a parameter:the cost of dealing with insurgents is proportional to their number.
Maximizing the government utility under these circumstances entails taking into account the costs of
civil unrest:
(C
w
1
)
Max
Q
Z
1
0
e
¡±t
[¼(p;S;Q) ¡B(Q) +¯B(Q) +U(S) ¡½(S)C(L(S;w))]dt
while maximizing the joint industry-government payo®s is equivalent to:
(C
w
2
)
Max
Q
Z
1
0
e
¡±t
[2¼(p;S;Q) ¡2B(Q) +¯B(Q) +U(S) ¡½(S)C(L(S;w))]dt;
subject to the familiar resource constraint.
The solutions to the two problems are given in Appendix A(6) and in a steady-state equilibrium the
following obtain:
(± ¡F
S
+
¹
H
S
)[¼
Q
+(¯ ¡1)B
Q
] = ¼
S
+U
S
¡½
S
C ¡½C
L
L
S
and (± ¡F
S
+
¹
H
S
)[2¼
Q
+(¯ ¡2)B
Q
] =

S
+U
S
¡½
S
C ¡½C
L
L
S
.
36
For a thorough exposition of these concepts in the context of resource con°icts see Homer-Dixon (1999),p.136.
37
See Renner (2002),p.40.
38
Alternatively,the ideal stock level could equal the long-run optimum as regulated by an honest government.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 13
Expressing the`bribing intensity'as B
Q
=
1
¯¡1
(
¼
S
+U
S
¡½
S
C¡½C
L
L
S
±¡F
S
+
¹
H
S
¡ ¼
Q
) from the ¯rst equation and
substituting into the second yields the modi¯ed augmented golden rule of renewable resource exploitation
under a corrupt government when con°ict is possible as:
(2.7) (F
S
¡
¹
H
S
) +
¯¼
S
+U
S
¡½
S
C ¡½C
L
L
S
¯¼
Q
= ±:
There are two opposite e®ects induced by the discontent-fueled unrest.On one hand,the cost of keeping
unrest under control induces the government to e®ectively regulate a higher level of the resource stock,
while on the other hand it has an incentive to draw the resource level further down,since this weakens
the local (potential rebel) workforce,thus decreasing the cost of discontent.The net e®ect of allowing for
the possibility that the locals may manifest their`grievance'in a costly way for the government depends
on the relative strength of the discontent and migration e®ects.
Comparing the resource stock implied by equation (2.7) with the one in (2.6) yields the following:
When the relative increase in discontent caused by a unit decrease in the level of resource stock exceeds
the corresponding cost savings brought about by migration,so that the net e®ect of depletion entails a
positive cost to the government or:¡½
S
C ¡½C
L
L
S
> 0,the marginal stock e®ect with con°ict is higher
than its no-violence counterpart.The intuition is the following:when local harvesters pose a credible
challenge by making it costly for the government to opportunistically deplete the resource,their implicit
policy in°uence increases.The ensuing regulation is then bound to be more conservationist.
Proposition 2:The specter of con°ict based on resource-related discontent of local harvesters under a
corrupt government has conservationist consequences,increasing the steady-state level of the resource stock
when the`discontent'e®ect dominates,i.e.for a low ideal stock level (S
¤
) and low outside opportunities
(w) and high maximum harvesting ability ¹q.The equilibrium stock decreases it if the`migration'e®ect
dominates,for opposite values of the parameters.
Proof:provided in Appendix A(7) and below.¤
Appendix A(7) shows the condition under which costly discontent increases the equilibrium level of
the resource stock takes the following form:G(S) = µS +(
¹qS
w
¡2)lnS ¡Á > 0,where µ = (
¹q
w
ln
¹q
w
+"),
Á = 2 ln
¹q
w
¡1 and"=
ln
¹qS
w
+1
S
¤
.While an analytical solution cannot be obtained,the following graph shows
via a numerical simulation that satisfying the inequality entails that the steady-state optimal stock level
exceeds a certain threshold level S'.
39
The discontent e®ect curve shifts up when ¹q is higher and when w is
lower.The migration e®ect shifts down when w and S
¤
are lower.Thus,it can be seen that the threshold
value S'is lower,thus easier to ful¯ll for higher µ,implying higher ¹q and lower w,as well as for higher",
which entails a lower S
¤
,since
lnS
¤
S
¤
decreases in S
¤
.
39
Notice that both functions in the ¯rst graph are constrained to be positive here.
14 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
stock
0
migration effect
discontent effect
S’
Depletion
effects
stock
M(S)
0
G(S)
M(S)
G(S)
S’ S
c
c
S
n/c
c
According to (2.6),the optimal stock level without con°ict is given by M(S) = (± ¡F
S
+
¹
H
S
)¯¼
Q
¡
¯¼
S
+U
S
= 0,while the optimal stock with con°ict arises at the intersection of M(S) and G(S),as shown
in the second graph.
40
The condition that the equilibrium stock level is`large enough'is also more likely
for a higher S
c
n=c
,which entails that the corruption coe±cient ¯,the ratio
r
K
and the price of the resource
good p are all low.
If the migration e®ect exceeds the discontent e®ect,the marginal stock e®ect and the equilibrium stock
are smaller when con°ict is possible.In this case the government has an incentive to purposefully deplete
the resource in order to drive out the potentially inconvenient locals.This occurs when the ideal stock level
and outside opportunities are high and/or when the local harvesters have a high harvesting ability.Under
full information,harvesters would consequently have less of an incentive to organize and mount pressure
on the government,knowing the result would be an acceleration of resource depletion and implicitly a
decrease of their utility.
Comparing the marginal stock e®ect implied from (2.7) to the one obtainable under an honest govern-
ment (2.1) opens the somewhat surprising possibility that:
Corollary 1:Under the specter of corruption-triggered con°ict,an opportunistic government can be
even more conservationist than the honest government.
41
This may occur when the government is not`too corrupt'relative to parameters:¯ < 1 ¡
½
S
C+½C
L
L
S
U
S
,
which after plugging in the functional forms introduced above is equivalent to ¯ ¡1 <
°
¹
h
.Such a situation
40
The simulation is based on logistic resource growth with intrinsic growth rate r and carrying capacity K:F(S) = rS(1 ¡
S
K
) and
the simplest industry pro¯t function ¼(p;Q;S) = (p ¡
c
S
)Q.
41
This result assumes con°ict is only possible when the government is opportunistic,or S
¤
= S
h
emphasizing the`grievance'aspect
of this rebellion:insurgency is legitimate only against a corrupt government.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 15
is facilitated when a high marginal cost of waging war for the government combines with a not too large
subsistence amount and relatively low corruption to yield a more conservationist policy outcome.
An important motivation of this paper was to understand the e®ect of resource policy and governance
quality on resource-based civil con°icts.Here the focus is on the e®ect of corruption on the likelihood of
con°ict,given`o±cial corruption'as related to the exploitation of natural resources is one of the main
justi¯cations of such rebellions.As mentioned above,civil unrest is called a civil con°ict when a certain
intensity of violence is reached.This is represented here by the expected cost of con°ict,denoted by
P(S) = ½(S)C(L(S;w)).The larger P,the more likely it is that a discontent-motivated civil con°ict
will occur in this setting.Di®erentiating this function with respect to the stock level of the contentious
resource S,the following obtains:
P
S
= ½
S
C +½C
L
L
S
.
While the locals'discontent increases as the level of stock goes down,the requirements of subsistence
consumption fosters emigration and weakens the rebellion threat,such that the probability of con°ict
increases in corruption for some levels of the resource stock and it decreases in corruption for others.In
this setting discontent-driven grievance-based resource con°ict arises when P
S
< 0,or when:
(2.8) ¡½
S
C > ½C
S
:
While (2.8) is an implicit relationship,it has an intuitive interpretation.When the increase in discontent
brought about by a one-unit decrease in stock level is more than the marginal cost savings associated with
the outward migration of potential rebels,the resulting likelihood of con°ict increases,situation which
is more likely when the equilibrium resource level is su±ciently high,which in turn requires parameters
to be as described in Proposition 2.The opposite occurs when the migration e®ect dominates,as the
government may be more able to keep the civil unrest under control and prevent it from erupting into a
full-blown con°ict.
As the corruption coe±cient only in°uences the probability of con°ict via the regulated level of deple-
tion,
42
its partial e®ect can be written as:
P
¯
= (½
S
C +½C
L
L
S
)S
¯
= P
S
S
¯
.
When the discontent e®ect dominates,corruption increases the likelihood of con°ict,since P
S
< 0 and
S
¯
< 0.
43
When the migration e®ect dominates,the impact of corruption of the likelihood of con°ict is
unclear,since P
S
> 0,but the sign of S
¯
is ambiguous.
42
Alternatively,one can imagine the quality of governance entering directly into a discontent function for the local harvesters,
although this would entail measurement issues related to the di®erence between real and perceived levels of corruption.
43
See Appendix A(8) for the derivation.
16 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
To analyze the manner in which the quality of governance a®ects this marginal e®ect of depletion on
the likelihood of resource-based civil con°icts,take the cross-partial derivative of P with respect to the
corruption coe±cient ¯:
(2.9) P

= (½
SS
C +2½
S
C
L
L
S
+½C
L
L
SS
¡µ
SS
)S
¯
:
Since when the discontent e®ect exceeds the migration e®ect the equilibrium level of the resource stock
is decreasing with the regulator corruption (S
¯
< 0),the impact of corruption on the marginal e®ect of
depletion on the likelihood of con°ict is a function of the sign of the bracketed expression in (2.9) which
equals P
SS
.Appendix (9) shows that P
SS
> 0 when the equilibrium level of stock exceeds S
00
,condition
which is more restrictive than the one stipulated in Proposition 2 above,yet is also more likely to be
satis¯ed for high ¹q and low w and S
¤
,as well as low ¯,p and
r
K
.Under such conditions,case in which
S
¯
< 0,the result that P

< 0 obtains.In words:
Corollary 2:The marginal e®ect of depletion on the likelihood of con°ict can decrease with corruption
as long as harvesting ability and carrying capacity of the environment are low,outside opportunities,ideal
stock level,governmental corruption,price of the product and intrinsic growth rate of the resource are
high.
This may help understand why the more corrupt places are not always the ones witnessing episodes of
resource-based civil con°ict,even though`governmental corruption'ranks high in the motivation list of
insurgents.
2.3.External pressure:trade and aid.
Despite their name,resource-based civil con°icts,are rarely purely`domestic',
44
but rather depend
on international trade,¯nance and aid networks.In the present model this international dimension is
represented by the possibility of trade measures against the regime and of international transfers or aid.
The optimal harvesting expression under a corrupt government under the threat of con°ict presented
in (2.7) allows for a few comparative statics that shed light on the impact of international pressure on
the quality of the domestic resource extraction regime.In particular,the e®ect of trade sanctions and
international aid,which are often the measures of choice to achieve such purposes can be readily derived.
In peace time,the large scale industry exports the good.Assume,as it is common in the literature,
45
that the e®ect of international sanctions against a regime that overexploits the environment is to reduce
instantaneously the producer price from commercializing the resource good.Then,the sole e®ect of the
resource-related sanctions here is to decrease the pro¯ts of the ¯rms ¼.Sanctions can be modeled as an
44
Le Billon (2000),points to globalization as responsible for the increased duration of such con°icts,due largely to easier ¯nancing,
arms trade and trade in illicit resources (p.3).
45
See,for instance,Damania and Barbier (2001).
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 17
ad-valorem import tari® for the resource good (¿),that can be varied from zero to in¯nity,to increase the
severity of the restriction.Then pro¯ts can be written as ¼(p
0
;S;H),where p
0
= p(1 ¡¿).
46
The goal of
the sanctions is then reached if the long-run equilibrium level of the resource stock is higher as a result of
their imposition.The following holds:
Proposition 3:A drop in pro¯ts caused by resource-targeted international trade sanctions will,ceteris
paribus,have a conservationist e®ect on the resource as long as the the conditions for P
SS
> 0 are satis¯ed.
Proof:see below and in Appendix A(10).¤
Di®erentiating expression (2.6) implicitly with respect to the price of the resource export p yields,upon
further transformation:
(2.10)
@S
@p
=
(± ¡F
S
+
¹
H
S
)¯¼
QS
¡¯¼
sp
¯¼
Q
(F
SS
¡
¹
H
SS
) +¯¼
SS
+U
SS
¡½
SS
C ¡2½
S
C
L
L
S
¡½C
L
L
SS
¡(± ¡F
S
+
¹
H
S
)¯¼
QS
:
As discussed in Appendix A(10),the denominator is negative if P
SS
> 0 equivalent to S > S
00
,due to
the concavity of the utility,pro¯t and resource growth functions.In addition,notice that when marginal
pro¯t is independent of the price (¼
sp
= 0),the numerator is positive,and therefore,the equilibrium
stock level is decreasing with price.The last assumption holds for the family of pro¯t functions for which
harvesting costs are not a function of price and harvested quantity is independent of the stock.The same
result holds,nevertheless,when,¼
sp
6= 0
47
but
¼
Qp
¼
Q
(¯¼
S
+U
S
) > ¯¼
sp
,where the latter inequality is shown
in Appendix A(10) to hold whenever P
S
< 0,or S > S
0
.
48
In this case,sanctions work to conserve the resource if they result in e®ective producer price reduc-
tions.Intuitively,sanctions increase the attractiveness of`investing'in the resource stock by lowering the
opportunity cost of conservation.In the political game,lower unit pro¯ts translate - since the optimal
bribe is truthfully re°ecting the marginal e®ect of increased quota on pro¯ts - into lower bribes,thus
inducing the regulation of lower harvesting quotas,and has an e®ect similar to a decrease in corruption.
In short,sanctions lower the attractiveness of the prize in the lobbying game.This ¯nding extends to this
particular political-economy framework the established result that in a one-species setting and where an
independent habitat value is not considered,international sanctions work to increase the long-run stock
of the resource.
49
As emphasized in the literature on the conservationist e®ect of international trade mea-
sures,ambiguous or perverse resource stock e®ects are only obtained when sanctions e®ectively change
the management regime to one of open access,or when the resource becomes a nuisance as an e®ect of
46
Note that,absent a domestic market for the resource good,an equivalent e®ect is observed when there is a negative shock to the
international price of the resource,or a hike in the extraction costs due,perhaps,to an increase in the price of an essential input.
47
In the present case the harvested amount depends on the stock level since it is equal to the regulated TAC which is determined
as the argument the maximizes the social welfare function and the joint regulator-client surplus.
48
Except for peculiar speci¯cations where production costs depend on ¯nal good prices.
49
This result appears,for instance,in Schulz (1996) and Barbier and Schulz (1997) and is contrary to a series of papers by Swanson
(1993),which claim sanctions are an ine®ective policy.
18 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
sanctions and the more pro¯table alternative use of its habitat renders over-depletion or even extinction
optimal.
International aid conditional on implementing a certain type of resource management is another possible
way to externally in°uence a country's resource extraction policy.
50
Recently,the global warming debate
has fostered the emergence of many such new initiatives which provide grants in return for conservation.
51
While unconditional transfers might have adverse rent-seeking e®ects on the host economy characteristic
to revenue booms,conditional aid may be successful in e®ecting positive change.Provided monitoring
capabilities exist,the transfer could be granted conditional on the level of the resource stock (T(S) with
T
S
> 0).Given the fact that the regime is corrupt,it can be assumed - based on case study evidence
- that only a part of the international transfer reaches its intended target.In other words,suppose the
government embezzles a fraction ® of this aid:
52
(C
T
1
)
Max
Q
Z
1
0
e
¡±t
[¼(p;S;Q) ¡B(Q) +¯[B(Q) +®T(S)] +U(S) ¡P(S)]dt
subject to
_
S = F(S) ¡Q¡
¹
H(S),where recall that P(S) = ½(S)C(L(S;w)),and P
S
= ½
S
C +½C
L
L
S
.
The truthful contribution and quota also solve:
(C
T
2
)
Max
Q
Z
1
0
e
¡±t
[2¼(p;S;Q) ¡2B(Q) +¯[B(Q) +®T(S)] +U(S) ¡P(S)]dt
subject to
_
S = F(S) ¡Q¡
¹
H(S).
The solution to this optimization problem is given in Appendix A(11) which provides the optimal
exploitation condition in the presence of resource management - targeted international aid:
(2.11) (F
S
¡
¹
H
S
) +
¯¼
S
+U
S
+®¯T
S
¡P
S
¯¼
H
= ±:
Again,the marginal stock e®ect is unambiguously increased by the transfer,since T
S
> 0 and ¯ > 1,
and so the transfer attains its conservationist goal.
Comparing this to the quota set by the corrupt government without international aid (2.6) and calcu-
lating the comparative statics yield the following:
Proposition 4:Conditional international aid leads to a higher equilibrium level of stock and the
magnitude of the reduction in harvesting is proportional to the`embezzlement'coe±cient ® (and can also
increase in the corruption coe±cient ¯).
Proof:provided in Appendix A(11).¤
50
Examples are:the Global Environmental Facility (GEF),which provides grants for nature protection,and debt-for-nature swaps,
whereby a conservationist agency and/or creditors of a (usually) developing country agree to sponsor conservationist policies.See also
Damania and Barbier (2001),p.16.
51
A 2007 initiative titled`Leaving Ecuador's Oil in the Ground,'seeks to attract international funding to avoid carbon emissions
and save the Yasuni Rainforest where`[at] least two indigenous tribes,the Tagaeri and Taromenane,maintain their traditional lifestyles
in voluntary isolation',according to the World Resources Institute.See http://www.wri.org/.
52
The rest is supposedly destined for a payo®-neutral use,for instance to pay o® a part of the country's foreign debt.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 19
The intuition for this result is simple.Aid directly decreases the attractiveness of resource exploitation
for the government.The fact that the transfer is partly appropriated by the government means that the
relative`value of conservation'rises in the corrupt government's objective,and so the equilibrium stock
level increases in the presence of aid conditional on the resource policy,with the increase being positively
related to regulator corruption.
53
It is straightforward to show that this result is not driven by particular assumptions related to the form
of the international transfer.Alternatively,the transfer T can be designed as a function of the regulated
harvesting:T(Q) where T
Q
< 0.The critical assumption is,again,that the international donors have
some monitoring capabilities in place for the resource policies set by the government:the quota Q is either
directly observable or can be inferred from the industry's ¯nancial statements.The government manages
to embezzle a fraction ® of the transfer and solves the following problem:
54
(C
T
1
)
0
Max
Q
Z
1
0
e
¡±t
[¼(p;S;Q) ¡B(Q) +¯[B(Q) +®T(Q)] +U(S) ¡P(S)]dt
subject to
_
S = F(S) ¡Q¡
¹
H(S).
The truthful contribution and quota also solve:
(C
T
2
)
0
Max
Q
Z
1
0
e
¡±t
[2¼(p;S;Q) ¡2B(Q) +¯[B(Q) +®T(Q)] +U(S) ¡P(S)]dt
subject to
_
S = F(S) ¡Q¡
¹
H(S) and the truthful resource exploitation policy rule satis¯es the following
relation,as Appendix A(11) further details:
(F
S
¡
¹
H
S
) +
¯¼
S
+U
S
¡P
S
¯¼
Q
+®¯T
Q
= ±;
where T
Q
< 0 and the marginal stock e®ect is again increased by the presence of the conditional aid.As
long as both trade sanctions and international aid increase the equilibriumlevel of the resource stock,they
contribute to a decreased equilibrium likelihood of con°ict.The preceding theoretical analysis,however,
points to several testable hypotheses which are discussed next.
3.An Empirical Analysis of Environmental Depletion and Governance-Based
Determinants of Civil Conflict:Revisiting the`Grievance'Hypothesis
This section investigates the hypothesized empirical relationship between the likelihood for a particular
country to be engaged in a civil war,basic economic indicators such as population and income levels,
and the nature of its governance and natural resource policies.The theoretical model of corruption and
renewable resources proposed above is employed to inform the empirical speci¯cation in an attempt to
53
Barbier and Schultz (1997) obtain a similar result in a setting where they consider habitat conversion (p.161).
54
The rest of the transfer has no further redistributive repercussions,as it is mandated to go towards a public good destination,e.g.
re-paying the country's external debt.
20 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
look into the potential impact of resource policies and quality of governance on the prevalence of domestic
con°ict.Apanel data set is used to estimate a reduced-formempirical relationship based on the theoretical
framework and thus to indirectly test the hypotheses advanced in the previous part.Due to a widely-
recognized under-theoretization of resource-con°icts,the choice of variables and speci¯cations in many
models has very much been ad-hoc.
3.1.Empirical implications.
The likelihood of resource-based con°ict is positively related to the local population level,which is
associated with the strength of the local harvesters in this model.In what follows,a country's population
is used as a proxy,based on the assumption that the larger the population,the more likely it is to have
such disenfranchised groups with a`critical mass'membership,while the government's capabilities are
more diluted.This is used for comparability with the literature.When rural population density is added
to all regressions,it is also positive and signi¯cantly related to con°ict incidence.However,this measure
also exhibits a strong negative correlation with the income measure.Con°ict is negatively related to the
level of outside opportunities,proxied for by income level and/or growth.While in the model the small
harvesters attracted by better options simply exit the interest group,this is represented in the data by
the notion that in a high income and/or high growth economy insurgency is less attractive as a career
choice,even in the presence of legitimate grievances.
Resource depletion and corruption increase the likelihood of con°ict when the parameters are such that
the resource is not too scarce in equilibrium,which in the model occur as the outside opportunities are
not very attractive for a high ability peasant population and the government's appeasement deterrence
capabilities are low.An advanced degree of scarcity induces a weakening of the locals and a lowering of
the probability that an open con°ict would ensue.The variables used to proxy for resource depletion (net
forest depletion and adjusted net savings) and governance quality (corruption and bureaucratic quality)
are discussed in more detail below.
The interaction between the natural resource depletion and the quality of governance is important:
corruption is less likely to cause violence in more depleted environments.While corruption fundamentally
increases the probability of violence,in some circumstances,corruption may in fact be instrumental in
allowing the government to appease the potential rebels weakened by migration.
The following presents the data,the empirical model and examines the validity of these ¯ndings in the
sample.
3.2.Data.To assemble the panel data set information on natural resources,civil con°icts,governance and
corruption,democracy and general macro-economic variables such as GDP level and growth or population
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 21
is required.Several di®erent sources are used for this purpose.A list of variables names,meaning and
their sources is provided in Appendix B,and a brief introduction is provided in the following paragraphs.
The natural resource data comes from the World Bank's Adjusted Net Savings database,which in
general provide measures of economics sustainability.The ANS project provides information since 1970
for a wide cross-section of countries on several indicators relevant to the main focus,such as:adjusted
net savings,net forest depletion,mineral depletion,energy depletion,gross and net national savings,
CO
2
pollution damage.
55
Following the model,a narrow focus lies primarily in the renewable resources
information,such as net forest depletion.
56
In a broad interpretation of the model,however,where
the environment is viewed as a generic renewable resource,the data on adjusted net savings is more
informative.
Civil con°ict data was obtained from the PRIO/Uppsala Armed Con°ict Dataset
57
which - compared
to previous e®orts - extends the con°icts set by lowering the casualty threshold necessary for an episode
of violence to qualify as a con°ict from 1000,in the Correlates of War (C.O.W) project,
58
to 25-battle
related deaths annually,while also keeping track of the intensity of the war.Thus,these data are much
more inclusive,capturing signi¯cantly more than just the major civil con°icts worldwide,over the period
1946-2005.The operational de¯nition of con°ict used in the database is`a contested incompatibility that
concerns government and/or territory where the use of armed force between two parties,of which at least
one is the government of a state,results in at least 25 battle-related deaths.'
59
The dataset also includes
inter-state con°icts,however,in accordance to the stated interest,only the internal ones are kept.
60
While
most of the previous empirical studies use the much more restrictive 1000 casualties per year threshold,
the position expressed here is the following:whereas that body of work looks actually to identify the
factors important in reaching that speci¯c intensity of hostilities,this data allows to test more directly for
determinants of civil con°ict emergence.Moreover,this inclusiveness is important for the identi¯cation
strategy,as explained later.
Reliable data about corruption and the general quality of governance are di±cult to collect,due to the
very nature of the phenomenon,premised on concealing its existence.Therefore all data sources available
55
Adjusted net savings are calculated as augmented standard national income accounting ¯gures de°ated by the Gross Na-
tional Income (GNI).Net forest depletion calculation is based of estimated depletion rents,`calculated as the rent on that amount
of extraction which exceeded the natural increment in wood volume.'More information is provided in the World Bank manual at:
http://siteresources.worldbank.org/INTEEI/1105643-1115814965717/20486606/Savingsmanual2002.pdf
56
Unfortunately,no similar data exist for ¯sh or wildlife.
57
See Gleditsch et al 2002.
58
First started by Singer and Small (1972),at the University of Michigan,transferred to Penn State in 2001.
59
See UCDP/PRIO Armed Con°ict Dataset Codebook,p.4.
60
Categories 3 and 4,for`internal armed con°ict'and`internationalized internal armed con°ict',the latter group including civil
wars that witness some form of external interference.
22 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
for a wide array of countries and time periods are not based on factual data,but rather on perceptions.
61
These perception-based corruption indicators are constructed with information frommultiple sources,and
so there are less chances of any systematic bias or measurement error.Among the most prominent in this
category of sources,one can count the Transparency International's Corruption Perception Index started
in 1995 with 41 countries gradually expanded to 158 in the present,the International Country Risk Guide
which looks speci¯cally at corruption in the political system,and the World Bank's Governance Indicators.
The source for corruption data used here is the International Country Risk Guide (ICRG) score,
62
which
refers speci¯cally to corruption in the political systemand to ties between business and politics.There are
a number of governance indicators available,such as:Government Stability,Socioeconomic Conditions,
Investment Pro¯le,Military in Politics,Religion in Politics,Law and Order,Ethnic Tensions,Democratic
Accountability,Bureaucracy Quality and Corruption,with the last two being the more relevant to the
present focus.In particular,the measure of corruption refers to`¯nancial corruption in the form of
demands for special payments and bribes...and suspiciously close ties between politics and business'.
63
As
with all ICRG indicators,higher values of variables cor and bqua signify better quality governance:less
corruption and higher bureaucratic quality,respectively.
Additional controls incorporated in the data set come from World Bank's Development Indicators for
variables such as:GDP levels and growth rates,population,agricultural products,food and ore exports,
the general level of inequality,unemployment,foreign aid and others.Additionally,data fromthe POLITY
IV project were used to account for the general level of democracy in a country in the form of a special
variable (polity2) designed speci¯cally for time-series analyses.
64
The ¯rst table details some important
variables used in the analysis.
Some correlations between important variables are presented in the second table.Two relationships
are particularly worth pointing out in this cross-correlations table.First,the ICRG indicators of corrup-
tion and bureaucratic quality,while positively correlated with the level of democracy in a country,are
not strong enough to actually instrument for one another,and can be justi¯ably inserted into the spec-
i¯cation as independent regressors controlling for di®erent aspects of a country's governance.Secondly,
61
Several studies analyze the distance between perception indicators and more factual-based measures of corruption.See Olken
(2006) for a recent example,where a local measure of corruption in road-building in Indonesia is compared with corruption perceptions.
Though interesting,this is not a signi¯cant problem in the present setting,where perceived mismanagement and embezzlement of public
funds can equally generate revolt.
62
International Country Risk Guide (Table 3B),C The PRS Group,Inc.,1984-Present.For more details see http:
==www:prsgroup:com.
63
Citation from the ICRG codebook A Business Guide to Political Risk for International Decisions,p.31.
64
For a thorough explanation of why a modi¯cation of the combined polity score is necessary,please refer to the original source:
Marshall and Jaggers,Polity IV Project:Data User's Manual at www.cidcm.umd.edu/inscr/polity.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 23
Variable Obs. Mean Std. Dev. Min. Max.

GDP growth 4238 3.287308 6.115611 -51.03086 106.2798
GDP per capita (PPP) 3906 8189.072 8563.14 466.1966 66702.59
Food exports 2798 25.23914 25.30327 .000238 99.11278
Fuel exports 2681 15.95988 25.86532 0 101.5603
Metal ore exports 2747 6.291353 11.86478 3.36e-06 88.81229
Gini coefficient 400 41.1678 10.80784 19.4 74.33
Aid 3267 8.251895 13.82001 -3.016838 242.2864
Aid per capita 3521 80.30627 215.12 -203.5889 2337.979
Population 4728 1.45e+08 5.95e+08 19700 6.44e+09
Rural population density 4013 555.3755 1121.809 0 13776.82
Subsidies and transfers 975 36.21356 21.67086 -.7346081 90.65208
Surface area 4536 3374.076 12867.36 .05 134000
Unemployment youth male 1178 16.37603 9.214946 1 69.2
Polity (democracy indicator) 3158 1.548132 7.255555 -10 10
Bureaucratic quality 2889 2.139171 1.20512 0 4
Control of corruption 2889 3.13674 1.382194 0 6.166667
Democratic accountability 2889 3.612924 1.646162 0 6
Conflict intensity 640 1.326563 .4693225 1 2
Peace before conflict 945 4.445503 9.4801 0 50
In conflict 945 .6772487 .4677761 0 1
ANS excluding PM10 2722 5.929805 13.77362 -108.63 42.36
ANS including PM10 1921 4.963566 13.8389 -109.65 41.67
ANS energy depletion 3582 4.217113 10.22142 0 79.12
ANS Gross National Savings 3292 18.27396 12.50415 -58.95 86.64
ANS mineral depletion 3596 .7258843 3.002218 0 56.95
ANS net forest depletion 3139 .5396974 1.559226 0 15.97
ANS Net National Savings 3289 7.919556 12.05643 -70.37 80.6
Table 1:Summary statistics - selected variables
Variables gdp gwth pop pol bqua cor anse ansi mid nfd end

Gdp per capita 1.00
Gdp growth -0.06 1.00
Population -0.06 0.14 1.00
Polity 0.46 -0.07 -0.07 1.00
Bureaucatic quality 0.78 -0.03 0.05 0.40 1.00
Corruption control 0.62 -0.06 -0.09 0.41 0.64 1.00
ANS excl. PM10 0.25 0.08 0.14 0.34 0.36 0.25 1.00
ANS incl. PM10 0.26 0.08 0.13 0.35 0.37 0.25 0.99 1.00
Mineral depletion -0.16 -0.02 -0.02 -0.00 -0.10 -0.06 -0.07 -0.07 1.00
Net forest depletion -0.25 0.05 0.00 -0.17 -0.28 -0.18 -0.13 -0.12 -0.03 1.00
Energy depletion -0.07 0.09 -0.00 -0.43 -0.13 -0.23 -0.64 -0.65 -0.08 -0.12 1.00
Table 2:Cross-correlations
the correlations between the indicators of depletion and corruption over the sample are also low,while
bearing the`right'signs.
3.3.Empirical method and results.Before looking at the main hypothesis related to the determi-
nants of civil con°ict,the following attempts to con¯rm - given the available data - another relationship
24 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
established in the theoretical part.The ¯rst proposition suggests a negative association between the qual-
ity of governance and the level of natural resource depletion.Indeed,as Table 3 illustrates,the lagged
bureaucratic quality is negative and signi¯cantly correlated with the speci¯c depletion indicator net forest
depletion nfd,and the relationship is robust to the inclusion of additional controls,such as income level
and growth,population or level of democracy.The same holds for the relationship between the generic
environmental indicator anse and the control of corruption.When inconflict and inconflict5 are
added in columns 3 and 6,governance quality is not signi¯cant any longer,although it maintains the hy-
pothesized sign.As a ¯rst piece of evidence for ruling out the possibility of reverse causation,our con°ict
indicator,inconflict5,which is a dummy that turns one when civil con°ict occurs within ¯ve years from
the observation year,
65
is not a signi¯cant determinant of depletion,while entering with a positive sign
for both environmental indicators.It appears that the anticipation of future con°ict does not encourage
curet-period over-exploitation in the sample.On the other hand,the lagged dummy for contemporaneous
con°ict inconflict is a signi¯cant determinant of net forest depletion,lowering exploitation and might
be interpreted to correspond to our ¯ndings in Proposition 2 in the preceding theoretical part.The two
con°ict dummies are not signi¯cant determinants of either nfd or anse when entered alone,without the
additional controls.
66
Variables Dependent Variable: nfd Dependent Variable: anse
(1)OLS (2) OLS (3) FE (4) OLS (5)OLS (6)FE

Bur.Quality (lag) -0.072*** -0.070*** -0.007
(0.016) (0.015) (0.046)
Corruption (lag) 0.733*** 0.919*** 1.057
(0.211) (0.219) (0.653)
Gdp/capita (lag) -0.013*** -0.032 0.242*** 0.041
(0.005) (0.025) (0.077) (0.443)
Population (ln,lag) 0.143** 0.376 2.135*** -0.877
(0.059) (0.254) (0.650) (4.306)
Polity (lag) 0.005** -0.007 0.085* -0.108
(0.003) (0.007) (0.047) (0.125)
Inconflict (lag) -0.178** -0.622
(0.072) (1.276)
Inconflict5 (lag) 0.074 2.006
(0.100) (1.753)
Constant 0.580*** -0.006 -0.985 3.113** -10.227*** 2.561
(0.124) (0.305) (1.343) (1.257) (3.466) (23.889)
Observations 2317 2150 591 2114 2029 536
Groups 131 121 51 122 117 47

Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
Table 3:Depletion and governance quality
65
This time horizon is chosen for compatibility with the previous literature.A non-binary alternative is intensitymax5,which
measures the maximum intensity of con°icts in a country during a period of 5 years from the observation year.However,the variability
of this indicator is also limited,as it is zero for cases where con°ict is absent,one for minor con°ict and two for large-scale civil war.
66
Table 2(3.1) in Appendix B presents some further robustness checks of these results.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 25
As for the determinants of civil con°ict,the interest is in analyzing the relationship between the inci-
dence of internal violence,resource depletion and quality of governance in a wide cross section of countries
during 21 recent years.The only correlates of civil war that are consistently robust in the empirical
literature are GDP per capita - displaying a negative relationship - and population levels - having a pos-
itive relationship - with both the onset and the incidence of domestic violence.The opportunity cost of
participating in an armed insurgency is higher in a high average-income country,while a larger popula-
tion per se may increase the likelihood that one particular group may rebel against the government,in a
territorial dispute or in a coup.Beyond these two factors,however,there is little agreement in the ¯eld.
Still,Collier and Hoe²er's stark initial claims identifying rebellion with a`quasi-criminal activity',based
on their ¯ndings that dependence on primary commodity exports are an important determinant in the
emergence of civil wars,while legitimate motives were dismissed,still seemto stimulate researchers.While
many papers concentrate on showing how by adopting di®erent de¯nitions,time periods or con°ict data
sets,their results cannot be replicated,few actually move forward and look for more (theoretically and
intuitively) satisfying explanations.
67
One chief reason for this may have been that the body of theoretical
literature to guide such empirical explorations is relatively slim.
Civil wars may have their origin in several di®erent social phenomena.One potential source of resource-
based con°icts is investigated here,namely discontent-driven rebellion.The following re¯nes the investi-
gation of a`grievance'based motive for internal violence.Previously,the proxies used for grievances were
broad and political-based:level of democracy and measures of ethnic and religious fractionalization.
68
Given the relatively large number of what appear to be resource-based con°icts out of the total civil
con°icts,an additional way one can think about resource exploitation and the quality of the governance
as factors leading a certain group to consider con°ict was suggested above.Resource depletion data such
as:net forest depletion,and adjusted net savings is used to proxy for the level of environmental exploita-
tion/scarcity in a narrow and a broad sense,respectively,and governance indicators such as:corruption
and bureaucracy quality to represent the quality of policy-making.Available additional controls include
the Gini coe±cient for income inequality,the percentage of unemployed young male as potential rebel
recruits,the polity score on the democracy-autocracy scale for the political regime in the country.
69
The
governance quality terms are also interacted with the depletion indicators.
67
Among the exceptions,one recent paper by Brunnschweiler and Bulte (2006) takes issue with the treatment of resource dependence
(share of primary exports in GDP) as resource abundance,and the exogenous treatment of what may essentially be an endogenously
determined variable.
68
On the latter,some studies argue that it could be also interpreted as preventing con°ict,since rebel recruitment is more di±cult
in more fragmented societies.
69
Elbadawi and Sambanis (2002) ¯nd that the level of democracy and the level of ethnic fragmentation are signi¯cant determinants
of civil con°ict in their sample.
26 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
The basic probit speci¯cation (without lags for simplicity) can be written as follows:
©
¡1
(Conflict
it
) = ¯
0

1
Income
it

2
Population
it

3
Gov:Quality
it

4
Depletion
it
+

5
(Gov:Quality ¤ Depletion)
it

6
Democracy
it

7
Z
it

i

it
where © is the standard normal cumulative density function,income
it
is either the level of GDP per
capita level (purchasing power parity) or GDP growth,Z includes other controls and i indexes the panels
(countries),t is time measured in years,º
i
is the panel-speci¯c,unobserved heterogeneity e®ect and ²
it
is
the error term.
70
Income per capita has been widely used as a control in regressions of civil con°ict.According to
the model,a high income (level and/or growth) represents a high level of outside opportunities,having
the e®ect of decreasing the numbers of potential rebels via migration,while population has an opposite
(positive) in°uence on the incidence of civil con°ict.
71
Also,expect the control of corruption to generally
decrease the likelihood a domestic con°ict ensues,while the depletion indicator should increase it.Recall
from the theoretical section that,for relatively depleted environments,the con°ict-inducing e®ect of
depletion decreases with corruption,which implies that the likelihood of con°ict should be lower,the lower
the ICRG corruption control score for these countries.Therefore,while the coe±cient for corruption is
expected to be negative,the interaction term between the corruption score and the depletion indicator
is hypothesized to be positive.The estimation method is random e®ects probit,and the results are
summarized in the following tables.
Since institutional variables are among the regressors,reverse causality is a potential problem.It could
be argued that a high degree of corruption and resource exploitation might be due to the presence of
con°ict.In an unstable environment,the government bureaucrats may have short appointment horizons
and might be attempting to make up for this`shortcoming'by accepting side-payments in exchange
for policy favours,and consequently the quality of governance decreases,exacerbating overexploitation.
This potential problem has been explored ¯rst in Table 3 above,where con°ict is shown not to be a
salient determinant of environmental depletion.To further account for this possibility,the analysis takes
advantage of the time structure of the data set and uses time lags for all right-hand-side variables.In
addition,notice that there is another advantage,besides comprehensiveness,to adopting a lower death-
threshold when de¯ning con°icts which is worth pointing out in this context.It could be argued that
70
©(z) ´
R
z
¡1
(2¼)
¡1=2
exp(¡x
2
=2)dx.Alternatively,the speci¯cation could be expressed as:Probability(Conflict
it
= 1) = ©(¯¢Z),
where ¯ and Z are,respectively,the vectors of coe±cients and regressors.
71
In an innovative paper that uses rainfall as an instrument for economic growth,Miguel et al.(2004) show that income growth
is also a signi¯cant determinant of domestic con°ict in Africa.Note that,although rainfall data exists for a wider cross-section of
countries,this identi¯cation strategy is not available here,since rainfall instruments for economic growth only in a particular subset of
agriculture-dependent economies where irrigation is lacking,such as Sub-Saharan Africa.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 27
attempts to rule out reverse causality by using one period-lagged regressors are not fool-proof,since adverse
consequences of the civil war can arise early,before the typical 1000-battle related deaths threshold is
reached.The particular con°ict database used here allows for an early detection of violence,so that the
likelihood of anticipatory behavior from the regulator is reduced.
Alternatively,the results might be spurious:perhaps the relationship captures the fact that both sets
of variables are trending,or perhaps there exists a factor that is not included in the regressions and
that in°uences both the incidence of con°ict and the governance quality variable.The latter source of
spuriousness is investigated below by instrumenting for the quality of governance.The former is ruled
out by including a time trend,and results do not change,except for slight increase in the signi¯cance of
income growth,population and polity variables.The ¯rst set of results are presented in Table 4,where the
two measures of governance quality (bureaucratic quality and control of corruption) and the two measures
of environmental depletion (adjusted net savings and net forest depletion) are entered alternatively.
As hypothesized,the income level and growth have discouraging e®ects on con°icts:when good al-
ternative options exist,participating in a rebellion is much less appealing.Also,the natural logarithm
of population level has a positive and strongly signi¯cant e®ect,con¯rming the hypothesis.The inter-
action term between the quality of governance variable and the depletion indicator is signi¯cant.This
statistical signi¯cance is interpreted to mean that the e®ect of resource depletion (forest in this case) on
the incidence of civil con°icts is dependent on the governance variable (here control of corruption and
bureaucratic quality),as hypothesized.Moreover,the relationship between the prevalence of con°ict and
the governance variable also appears to exhibit the characteristics proposed in the theoretical section.
While the (previous period) control of corruption and bureaucratic quality are negative and strongly sig-
ni¯cant,the interaction with the indicator of net forest depletion is positive and signi¯cant and with the
adjusted net savings indicator (opposite of depletion) negative and signi¯cant.Calculating the marginal
e®ects of the quality of governance and depletion variables con¯rm that the adjusted net savings and the
control of corruption both decrease the likelihood of con°ict in column 1,while the net forest depletion
increases the likelihood of con°ict and the bureaucratic quality decreases the likelihood of con°ict in col-
umn 3.
72
These results provide con¯rmation of the hypothesis derived in the theoretical section:while
a poor governance quality is positively correlated with the incidence of civil con°icts,its e®ect is lower
given a high level of environmental depletion.
73
In the fourth column,the e®ect of regime longevity has
a negative sign,corresponding perhaps to a larger government capability to prevent con°ict,yet it is not
72
The marginal e®ects on con°ict evaluated at means are -0.0001(0.0004) for anse and -0.031(0.0012) for cor,respectively 0.029(0.017)
for nfd and -0.034(0.015) for bqua,where parentheses contain the associated standard errors.
73
Notice the anse term is positive,however,when entered without the interaction term,it has the intuitive negative sign.When
nfd is entered alone,it is a positive and signi¯cant determinant of con°ict.
28 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
Variables Dependent variable: inconflict5
(1)PanelProbit (2)PanelProbit (3)PanelProbit (4)PanelProbit (5)PanelProbit (6)PanelProbit

Gdp / cap (lag) -0.040* -0.041* -0.017 0.015 -0.002 -0.041
(0.024) (0.025) (0.027) (0.034) (0.028) (0.025)
Gdp growth (lag) -0.041*** -0.040***
(0.014) (0.012)
Population (ln,lag) 0.580*** 0.579*** 0.578*** 0.685*** 0.690*** 0.609***
(0.128) (0.122) (0.127) (0.127) (0.137) (0.125)
Corruption (lag) -0.361*** -0.409*** -0.419***
(0.075) (0.085) (0.086)
Bur.Quality (lag) -0.504*** -0.513*** -0.506***
(0.100) (0.104) (0.102)
Adjusted Net Sav (lag) 0.038** 0.027** 0.027**
(0.015) (0.012) (0.011)
Net forest depletion (lag) -0.110 -0.036 -0.148
(0.158) (0.097) (0.163)
Polity (lag) -0.005 0.003 -0.003 -0.021 -0.019 0.004
(0.014) (0.014) (0.013) (0.015) (0.014) (0.014)
Corruption*anse (lag) -0.012***
(0.004)
Corruption*nfd (lag) 0.107* 0.124**
(0.058) (0.060)
Bur.Quality*anse (lag) -0.017** -0.016**
(0.007) (0.007)
Bur.Quality*nfd (lag) 0.183**
(0.076)
Regime longevity -0.005
(0.005)
year -0.110*** -0.112*** -0.104*** -0.099*** -0.101*** -0.113***
(0.012) (0.011) (0.011) (0.012) (0.012) (0.011)
Constant 215.459*** 221.120*** 204.569*** 194.019*** 196.823*** 221.387***
(23.161) (22.325) (21.953) (24.608) (23.220) (22.470)
Observations 2128 2266 2266 2017 2127 2262
Groups 117 121 121 117 117 121

Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
Table 4:Civil con°ict incidence
statistically signi¯cant.When the dependent variable is the binary variable onset instead,de¯ned as a
binary variable equal to one when there is a con°ict in the current period conditional on previous period
peace,the signs are generally maintained,but the signi¯cance is weakened,as one might expect by only
considering contemporaneous con°icts.
74
When running a panel logit,results are similar.
75
When employing a ¯xed e®ects panel logit instead,
76
many groups drop out (the sample is signi¯cantly reduced,due to the existence of all positive or negative
outcomes within groups),
77
but the results do not change:the democracy level variable polity is no
longer signi¯cant (since it varies little in time),but the corruption variable and the interaction term
74
See Table 3(5.2) in Appendix B.
75
See Table 7(5.3),Appendix B.
76
Same table,columns 5,6.
77
The panel ¯xed e®ects logit estimator is based on the probability of observing a positive outcome in the panel.Consequently,
groups with all-positive outcomes have a conditional probability of being observed equal to one,and so are uninformative.See Gould,
W.1999.`Within-group collinearity in conditional logistic regression.'In Stata FAQs.College Station,TX:Stata Corporation.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 29
that constitutes the principal focus remain signi¯cant at 5% and have the hypothesized sign.The above
regression was also ran in speci¯cations featuring other dimensions of resource depletion.When using other
indicators from the Adjusted Net Savings database,such as the Minerals Depletion,Energy Depletion,
most of the results survive.This seems to suggest that resource depletion in general,is a signi¯cant
determinant of con°ict in the sample.
78
The paper also attempts to replicate some of the Collier and
Hoe²er's ¯ndings using our data.However,none of the other primary commodity exports (agricultural
commodities,fuel,food,minerals and ores) seem to have a signi¯cant in°uence on the incidence of civil
war in the sample.
79
Another potential source of endogeneity bias is the following.Perhaps some of the right-hand side
variables are jointly determined:low quality governance,or institutions may be the result of high levels
of income per capita.This is the issue to which the analysis turns next.
80
Studying the role of institu-
tions is di±cult,given their plausible endogeneity to almost any conceivable model speci¯cation.Good
instrumental variables are not easily available,and the practice has been to rely on the few ones coming
mostly from the cross-country development literature.La Porta et al (1998,1999) establish that the legal
origin of the country is a salient determinant of the legal protection of investors and thus of economic
development.
81
Acemoglu,Johnson and Robinson (2001) show that settlers mortality,available for 64
countries,is a good indicator for the type of institutions created by Europeans in the colonies,especially
with regards to property rights protection.Bockstette,Chanda and Putterman (2002) present data on
the antiquity of the state and argue for its usefulness as an instrument for the quality of institutions,if
one is ready to accept a long-run learning-by-doing process in governance.While the strength of property
rights is,undeniably,crucial for development,it is less of an issue in view of the theoretical framework,
where we assume property rights are reasonably enforced.In the ¯rst stage data on state antiquity and
legal origin is used,which are the more reasonable instruments for the quality of institutions in the sense
used here,while also implying the smallest restriction of the sample of countries.
An instrumental variable approach is implemented in an IV-2SLS framework,which is preferable to a
panel probit in the presence of endogeneity bias.This approach was previously followed in Miguel et al
(2004).
82
Moreover,the OLS results can be given an easier interpretation.In the ¯rst stage regress the
78
See Appendix B,Table 8(5.4).
79
See Appendix B,Tables 9 and 10.
80
The quality of institutions has an overwhelming impact on the process of economic development,taking precedence over classical
factors such as geography and trade,as shown in Rodrik,Subramanian,and Trebbi (2002).Presumably,they also are an important
ingredient in the prevalence for domestic violence in society.
81
The categories are British,French,Scandinavian,German and Socialist origin of the legal system.See La Porta et al.(1999).
82
See also Wooldridge (2002).Miguel et al (2004) argue that the Rivers and Vuong (1998) two-stage conditional maximumlikelihood
estimator method designed for cross-sectional analysis`requires strong speci¯cation assumptions'to be translatable to panel data (p.
738).The same idea is expressed in Elbadawi and Sambanis (2002),p.327.
30 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
quality of governance indicators on the instruments and additional controls.The antiquity of the state
(statehist05) and the legal origin (legor) indeed emerge as preferred instruments and their quality
is better for the case of bureaucratic quality.Then run the IV regression of con°ict incidence on the
instrumented endogenous and hypothesized exogenous regressors and the results are presented in Table
5.Notice that most of the results that were previously emphasized survive in the regressions where the
quality of governance is instrumented for:income is negative and signi¯cant,population is positive and
signi¯cant,the quality of governance is negative and signi¯cant,although the interaction e®ect which is
no longer signi¯cant.Calculating the marginal e®ects of depletion and of quality of governance on the
likelihood of civil con°ict con¯rms again the hypotheses.
83
Variables Dependent variable: inconflict5
(1)Panel Probit (2) Probit (3) Panel OLS (4)IV (5)Panel OLS (6)IV

Gdp growth -0.025 -0.010 -0.003 -0.003 -0.006 -0.004
(0.010)** (0.005)* (0.001)*** (0.001)** (0.001)*** (0.002)
Population (ln,lag) 0.574 0.303 0.056 0.060 0.072 0.076
(0.129)*** (0.023)*** (0.019)*** (0.018)*** (0.019)*** (0.032)**
Bur. Quality (lag) -0.538 -0.368 -0.045 -0.071
(0.097)*** (0.033)*** (0.010)*** (0.039)*
Corruption ctrl (lag) -0.043 0.212
(0.008)*** (0.174)
Net forest deplet. (lag) -0.033 0.118 -0.006 -0.081
(0.098) (0.046)** (0.013) (0.121)
Adjusted Net Sav (lag) 0.003 0.064
(0.002)* (0.031)**
Bur. Quality*nfd (lag) 0.193 0.061 0.016 0.117
(0.077)** (0.034)* (0.007)** (0.140)
Corruption*anse (lag) -0.001 -0.022
(0.000)* (0.011)**
Polity (lag) -0.005 0.003 -0.005 -0.005 -0.004 -0.014
(0.014) (0.005) (0.002)*** (0.002)** (0.002)** (0.008)*
Year -0.105 -0.044 -0.009 -0.010 -0.012 -0.009
(0.011)*** (0.005)*** (0.001)*** (0.001)*** (0.001)*** (0.005)*
Constant 205.507 86.160 18.705 19.845 23.447 16.980
(21.896)*** (10.782)*** (2.068)*** (2.495)*** (2.237)*** (10.777)
Observations 2289 2289 2289 2160 2129 2024
Groups 124 124 124 116 117 111

Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
Table 5:Instrumental variable approach
In addition to being correlated with the endogenous variable - here the quality of governance - valid
instruments have to also satisfy the exclusion restriction,i.e.they have to not be signi¯cant determinants
of con°icts on their own.While this is plausible for the antiquity of the state,which likely does not
83
The marginal e®ect (and standard errors) of net forest depletion (nfd) in column 3 is 0.01(0.005) and the marginal e®ect of
bureaucratic quality (bqua) is -0.015(0.003),while in column 5 the marginal e®ect of adjusted net savings (anse) is not signi¯cant
0.0001(0.0003),and the marginal e®ect of corruption control (cor) is -0.018(0.003).The signs of the marginal e®ects are generally
preserved but signi¯cance is lost in the IV regressions in columns 4 and 6.
CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES 31
have an impact on civil con°icts directly or through the other regressors than the one it instruments for,
legal origin may determine the economic regressors.However,when only state antiquity is used as an
instrument,the results do not change,apart from some decreases in the level of signi¯cance.
84
In the admittedly simplistic theoretical model,only resource policy can lead to con°ict.For this reason
the empirical analysis needs to control for other possible explanations.Many empirical investigations of
civil con°icts include indicators of society fractionalization,be it ethnic,religious or linguistic.Table 4 in
Appendix B shows that our principal results are robust to controlling for the degree of heterogeneity in the
population,although ethnic and linguistic fractionalization are positively correlated with the incidence of
con°icts.Additional robustness checks are also reported in Appendix B.
4.Summary and conclusions
Natural resources can be a powerful motivator of social and political change.The issue of resource-based
civil con°icts is a growing topic in both political science and economic literatures,and it has so far been
approached from various angles.This paper looks at the e®ect of o±cial corruption on the exploitation
of renewable resources and then on the likelihood for the emergence of con°ict,as it attempts to derive
testable implications.The study builds on classical political economy models like Grossman and Helpman
(1994) and looks at a dynamic externality in the formof renewable resource harvesting and three economic
actors:the opportunistic government,the resource corporations that are in a position to in°uence the
harvesting quota by bribing the regulator,and the peasant harvesters that live o® the resource.
For tractability reasons,the analysis focuses on the steady-state equilibrium,from which transitional
harvesting can be inferred.It is shown in a framework where con°ict is not a factor,that the opportunistic
government is less conservationist than an honest one.Resource depletion sanctioned by an opportunistic
government on one hand increases locals'discontent and the government's appeasement or deterrence
costs.On the other hand,in conjunction with the outside opportunities,depletion may weaken the
peasant (and potential rebel) workforce.The shadow of con°ict can be conservationist when the ¯rst
e®ect dominates,by increasing the implicit political in°uence of the small harvesters.The probability
of resource-based con°ict may increase or decrease,according to the discontent e®ect and the migration
e®ect,which in turn depend on the parameters of the system.International sanctions aimed at improving
the domestic environmental performance and which result in decreasing ¯rms'pro¯ts are shown to work
under reasonable conditions,while the e®ectiveness of environmentally-targeted international aid may
depend on corruption in a counter-intuitive way.
84
These results are also reported in Appendix B.
32 CORRUPTION AND CONFLICT IN THE MANAGEMENT OF ENVIRONMENTAL RESOURCES
It is then argued that the relationship between governmental corruption,the environmental outcome,
and the political outcome is complex.Moreover,without positing speci¯c functional forms that will
put a dent in generality,this issue is primarily empirical.The main hypotheses put forth are that a
higher population level and a lower income level increase the chances for violence and that while depletion
increases the probability of con°ict in`not too depleted environments',the marginal e®ect of depletion on
the likelihood of resource-based con°ict may decrease in corruption.
These hypotheses constitute the starting point for the empirical application of the paper,which tests
primarily the importance of corruption and other indicators of governance quality in conjunction with the
level of depletion of natural resources,as factors explaining the incidence of civil con°ict.A panel data
set containing a large cross section of countries during 20 recent years is used to show that corruption
and depletion and their interaction appear,indeed,to signi¯cantly in°uence the chance for civil violence,
lending support to the`grievance'motive of con°ict.This seems to suggest that it is not inconceivable,
at least in some cases,that economic policy grievances are more than mere justi¯cations used by rebels
to mask their real objectives.In contrast to the appropriation channel which o®ers little in terms of
policy advice,the resource policy mechanism yields the following policy prescription:a more inclusive
resource policy in conjunction with better overall economic conditions are likely to decrease civil con°ict
incidence.The implicit uniformizing nature of an analysis like the one above prevents one to draw sharper
conclusions.Case studies can be useful for more in-depth analysis of any particular instance of con°ict,
given the multitude of factors,their interactions and the incentives for misrepresentation of their true
nature that concur and may lead to eruption of violence.Notice,however,that little in the analysis above
applies strictly to resource-rich countries,unlike the empirical studies premised on the`greed'assumption.
Frustration with almost any type of government policy could potentially be accommodated in a similar
framework.
Given the explicit modeling choices made in both the theoretical and the empirical sections,the model
proposed here does not claim to be a universal explanation of civil con°ict.Nor is the empirical exercise
solid proof that civil con°ict is determined by resource depletion grievances when the policy channel is
by assumption unavailable or insu±cient to achieve a political balance in society.Rather,this is an
attempt to,on one hand,suggest that economic analyses of con°ict do not have to be premised on a priori
dismissing legitimate causes but can accommodate them,and on the other hand,to point at the fragility
of what still constitutes`conventional wisdom'in the empirical studies of civil con°ict.
APPENDIX A:
(1).The current value Hamiltonian for the honest government's problem can be written as:
H = ¼(p;S;Q) +U(S) +¹[F(S) ¡Q¡
¹
H(S)] and the ¯rst order conditions are the following:
dH
dQ
= ¼
Q
¡¹.
Notice that for speci¯c pro¯t functions that are multiplicatively separable in Q and take the form
¦(p;S)Q this condition does not depend on the control variable,and so the solution is reached by taking
the most rapid approach path (MRAP) as follows:if
dH
dQ
> 0 by setting harvesting at the maximum
level Q = Q
max
,where Q
max
is the exploitation resulting from employing the maximum e®ort available
(bounded perhaps by the available industry labour force and/or technological parameters),while if
dH
dQ
< 0
by setting harvesting at the minimum level Q = Q
min
(which can be zero or the break-even point for the
¯rms).This is allowed until the shadow price of the resource ¹ respectively increases or decreases to equal
the unit pro¯t.The other optimality relations are derived from:
dH
dS
= ±¹ ¡ _¹ = ¼
S
+U
S
+¹(F
S
¡
¹
H
S
)
dH

=
_
S = F(S) ¡H ¡
¹
H(S):
while the following transversality condition has to hold:
lim
t!1
e
¡±t
¹(t)S(t) = 0.
Imposing the conditions of zero-growth of the state and co-state variables in a steady-state equilibrium
and assuming the transversality condition holds,we obtain:
(F
S
¡
¹
H
S
) +
¼
S
+U
S
¼
Q
¡u
S
= ±:
To determine the transitional path dynamics,take time-derivative of the ¯rst condition to obtain:
_¹ = ¼
QQ
_
Q+ ¼
QS
_
S which can be substituted into the second condition as follows:(± ¡ F
S
¡
¹
H
S

Q
¡
¼
QQ
_
Q¡¼
QS
[F(S) ¡Q¡
¹
H(S)] = ¼
S
+U
S
.The
_
H = 0 locus can then be determined as:
H =
¼
S
+U
S
¡(± ¡F
S
¡
¹
H
S

Q
¼
QS
|
{z
}
+(F(S) ¡
¹
H(S)),where the sign of its slope (
dH
dS
j
_
H=0
) is ambiguous.
Notice that setting the numerator of the fraction to zero,which is our golden rule above,is equivalent to
imposing the steady-state harvesting condition.Transitional harvesting is decreasing towards the steady-
state value when the fraction (its numerator) is positive,and increasing when it is negative.¤
(2).The current value Hamiltonian of the corrupt government's problem can be written as:
H = ¼(p;S;Q) +(¯ ¡1)B(Q) +U(S) +¹[F(S) ¡Q¡
¹
H(S)].
Note,again,that the government is opportunistic for values of ¯ above 1.According to the maximum
principle,the ¯rst order conditions for an interior solution are given by:
dH
dQ
= 0 = ¼
Q
+(¯ ¡1)B
Q
¡¹
dH
dS
= ±¹ ¡ _¹ = ¼
S
+U
S
+¹(F
S
¡
¹
H
S
)
dH

=
_
S = F(S) ¡Q¡
¹
H(S);and we assume the transversality condition to hold.
34
Taking the time derivative in the ¯rst condition above yields:_¹ = ¼
QQ
_
Q+(¯ ¡1)B
QQ
_
Q+¼
QS
_
S and
plugging it into the second condition:
(± ¡F
S
+
¹
H
S
)[¼
Q
+(¯ ¡1)B
Q
] ¡¼
QQ
_
Q¡(¯ ¡1)B
QQ
_
Q¡¼
QS
_
S = ¼
S
+U
S

(3).The problemof maximizing the joint industry-government payo®s is solved in a very similar fashion:
Use current value Hamiltonian H = 2¼(p;S;Q) +(¯ ¡2)B(Q) +U(S) +¸[F(S) ¡H ¡
¹
H(S)] to get
¯rst order conditions:
dH
dQ
= 0 = 2¼
Q
+(¯ ¡2)B
Q
¡¸
dH
dS
= ±¸ ¡
_
¸ = ¼
S
+U
S
+¸(F
S
¡
¹
H
S
)
dH

=
_
S = F(S) ¡Q¡
¹
H(S):
Together these conditions imply:
(± ¡F
S
+
¹
H
S
)[2¼
Q
+(¯ ¡2)B
Q
] ¡2¼
QQ
_
Q¡(¯ ¡2)B
QQ
_
Q¡2¼
QS
_
S = 2¼
S
+U
S

(4).(i) A higher discount rate ± means that in order for equation (2.6) to hold,the equilibrium stock
of the resource will adjust to a lower level,corresponding to a higher net growth rate F
S
¡
¹
H
s
and a
lower marginal stock e®ect (MSE),both evaluated at the optimal stock level.A more impatient corrupt
government will regulate harvesting so that the resource stock settles at a lower long-run level.
(ii) To show Lemma 3,assume we have two functions of the stock of the same resource MSE
1
(S) and
MSE
2
(S),where MSE
2
(S) > MSE
1
(S);8S 2 [0;K].This is then true,in particular,for S = S
¤
2
:
MSE
2
(S
2
) > MSE
1
(S
2
) (¤).
We also know that the following two relations governing optimal exploitation are veri¯ed at the optimum
levels of stock S
¤
1
and S
¤
2
:
F
S
(S
¤
1
) +MSE
1
(S
¤
1
) = ± and
F
S
(S
¤
2
) +MSE
2
(S
¤
2
) = ±,where the resource dynamics and the discount rate are common across the
two problems.Assume,by contradiction,that S
¤
2
< S
¤
1
.It follows that F
S
(S
¤
1
< F
S
(S
¤
2
) and then for
the above equations to hold concomitantly,we need MSE
1
(S
¤
1
) > MSE
2
(S
¤
2
) (¤¤).Put together,the
two starred relations imply that MSE
1
(S
¤
1
) > MSE
1
(S
¤
2
),which further implies - since the functions
representing the marginal stock e®ects under the di®erent policy regimes are decreasing in the level of
stock - that S
¤
2
> S
¤
1
,which is a contradiction.¤
(5).It is easy to see that MSE
c
n=c
< MSE
h
n=c
(where the subscripts refer to`no-con°ict'and the
superscripts to government corruption) i®
¯¼
S
+U
S
¯¼
H
<
¼
S
+U
S
¼
H
.This is true in all cases,given the assumption
that ¯ > 1.
Alternatively,di®erentiate (2.6) with respect to the corruption coe±cient to obtain:
¯
2
¼
2
Q
(F
SS
¡
¹
H
SS
)S
¯
+S
¯
[(¯¼
SS
+U
SS
)¯¼
Q
¡(¯¼
S
+u
S
)¯¼
QS
] +¯¼
Q
¼
S
¡(¯¼
S
+U
S

Q
= 0,
35
which implies:
S
¯
=
U
S
¼
Q
¯
2
¼
2
Q
(F
SS
¡
¹
H
SS
)+(¯¼
SS
+U
SS
)¯¼
Q
¡(¯¼
S
+u
S
)¯¼
QS
< 0,
where the inequality derives from the concavity of functions F,¼ and u.¤
(6).The current value Hamiltonian for the government utility maximization problem when con°ict is
possible is:
H = ¼(p;S;Q) + (¯ ¡ 1)B(Q) + U(S) ¡ ½(S)C(L(S;w)) + ¹[F(S) ¡ H ¡
¹
H(S)] and the ¯rst order
conditions are:
d
H
dQ
= 0!¼
Q
+(¯ ¡1)B
Q
= ¹
dH
dS
= ±¹ ¡ _¹ = ¼
S
+U
S
¡½
S
C ¡½C
L
L
S
+¹(F
S
¡
¹
H
S
)
dH

=
_
S = F(S) ¡H ¡
h:
Take derivative with respect to time in the ¯rst condition above and plug into the second to obtain:
(± ¡F
S
+
¹
H
S
)[¼
Q
+(¯ ¡1)B
Q
] ¡[¼
QQ
+(¯ ¡1)B
QQ
]
_
Q = ¼
S
+U
S
¡½
S
C ¡½C
L
L
S
In the steady state:(± ¡ F
S
+
¹
H
S
)[¼
Q
+ (¯ ¡ 1)B
Q
] = ¼
S
+ U
S
¡ ½
S
C ¡ ½C
L
L
S
and this implies:
B
Q
=
1
¯¡1
[
¼
S
+U
S
¡½
S
C¡½C
L
L
S
±¡F
S
+
¹
H
S
¡¼
Q
].
Similarly,the joint industry-government surplus maximization yields in the steady-state:
(± ¡F
S
+
¹
H
S
)[2¼
Q
+(¯ ¡2)B
Q
] = 2¼
S
+U
S
¡½
S
C ¡½C
L
L
S
.
Plug in B
Q
fromabove to get the steady state equilibriumpolicy rule for renewable resource exploitation
when con°ict is possible:F
S
¡
¹
H
S
+
¯¼
S
+U
S
¡½
S
C¡½C
L
L
S
¯¼
H
= ±:¤
(7).Compare this to the case without con°ict and with a corrupt government in (2.6).Notice that
the marginal stock e®ect with con°ict is larger than the marginal stock e®ect without con°ict if ¡½
S
C ¡
½C
L
L
S
> 0 or if
¡½
S
½
>
C
L
L
S
C
.Given the expressions for ½,C and their derivatives with respect to stock
provided in the text,this becomes equivalent to ln
¹qS
w
(¹q ¡
w
S
) >
w
S
¡
w
S
¤
+
wln
¹qS
w
S
¡
wln
¹qS
¤
w
S
¤
,which is
further equivalent to
(
¹qS
w
¡2) ln
¹qS
w
¡1
S
> ¡
ln
¹qS
w
+1
S
¤
= ¡",(
¹q
w
ln
¹q
w
+")S + (
¹qS
w
¡ 2) lnS ¡ 2 ln
¹q
w
¡ 1 > 0.
Denoting by µ the coe±cient of S and with Á the constant,the previous inequality can be written:
G(S) = µS +(
¹qS
w
¡2)lnS ¡Á > 0.
Function M(S) is de¯ned in the text based on equation (2.6) as being equal to (± ¡F
S
+
¹
H
S
)¯¼
Q
¡
¯¼
S
+U
S
.Substituting the given functional forms,it becomes:
M(S) =
4pr
K
S
4
+(±p ¡2pr ¡
5cr
K
)S
3
+c(3r ¡±)S
2
+
¹
h(2pw ¡c¹q ¡
1
¯
ln
¹q
w
)S ¡
¹
h
¯
S lnS ¡c
¹
hw.
(8).Di®erentiating (2.7) implicitly with respect to ¯ yields:(¼
S
+ ¯¼
SS
S
¯
+ U
SS
S
¯
¡ ½
SS
CS
¯
¡
½
S
C
L
L
S
S
¯
¡½
S
C
L
L
S
S
¯
¡½C
L
L
SS
S
¯
)¯¼
H
¡(¯¼
S
+u
S
¡½
S
C¡½C
L
L
S
)(¼
H
+¯¼
HS
S
¯
) = ¡F
SS
S
¯
¯
2
¼
2
H
:
This implies:S
¯
=
¼
H
(u
S
¡½
S
C¡½C
L
L
S
)
(¯¼
SS
+U
SS
¡½
SS
C¡2½
S
C
L
L
S
¡½C
L
L
SS
)¯¼
H
+F
SS
¯
2
¼
2
H
¡(¯¼
S
+u
S
¡½
S
C¡½C
L
L
S
)¯¼
HS
< 0,if S >
36
S
0
and S > S
00
,since under the ¯rst condition the numerator is positive as P
S
< 0,while under the second,
the denominator is negative,as P
SS
> 0 (see below).¤
(9).P(S) = ½(S)C(L(S;w)) and it was shown above that P
S
= ½
S
C + ½C
L
L
S
< 0 if and only if
S > S
0
.Calculate the condition P
SS
= ½
SS
C +2½
S
C
L
L
S
+½C
L
L
SS
> 0 by using the functional forms
provided above as follows:(2 ln
¹qS
w
¡ 1)(¹q ¡
w
S
) ¡
2w
S
ln
¹qS
w
¡ 2w(
1
S
¡
1
S
¤
) ¡ 2w(
ln
¹qS
w
S
¡
ln
¹qS
¤
w
S
¤
) > 0,
2(
¹qS
w
¡3)(ln
¹q
w
+lnS)¡1 > ¡aS,where a =
2
S
¤
(1+ln
¹qS
¤
w

¹q
w
which further yields (
2¹q
w
ln
¹q
w
+a)S+2(
¹qS
w
¡
3) lnS ¡ (6 ln
¹q
w
+ 1) > 0.Denoting with b the coe±cient of S and with c the constant,this becomes:
J(S) = bS +2(
¹qS
w
¡3) lnS ¡c > 0.The analysis is now similar to the one in section (7) above,and the
graphical solution points to the fact that P
SS
> 0 if and only if S > S
00
,which is lower for higher ¹q and
lower w and S
¤
.
stock
0
G(S’)=0
S’’
J(S’’)=0
S’
stock
S’ S”
P(S)
0
P
S
>0
P
SS
<0
P
S
<0
P
SS
<0
P
S
<0
P
SS
>0
(10).We need the numerator of the fraction in (2.12) to be positive.The result holds for all pro¯t
functions for which the price derivative depends on the level the resource only via quantity,or ¼
pS
= Q
S
.
Under this reasonable assumption,the result is general.To see this,notice that the numerator in (2.11)
is positive if and only if

@S
+
U
S
¯
¡ P
S
>
¼
Sp
¼
Qp
¼
Q
.For S > S
0
(thus P
S
< 0),a su±cient condition
for this last inequality to hold is that
¼
Sp
¼
Qp
¼
Q
<

@S
.When ¼
Sp
= ¼
pS
= Q
S
and ¼
Qp
= ¼
pQ
= 1
(e.g.¼ = [p ¡ c(S)]Q(S)),this su±cient condition is equivalent to Q
S
¼
H
<

@S
.This is true,since

@S
= ¼
S

Q
Q
S
:the partial e®ect of S on pro¯ts is composed of a direct e®ect that works via lower costs,
and another indirect e®ect that works through a higher steady state harvesting quota.The denominator
of the fraction in (2.12) is negative when S > S
00
,thus P
SS
> 0.The graph below plots the probability of
con°ict as a function of the resource stock,where S
0
< S
00
and lim
S!0
P(S) = ¡1,which obtains since
lim
S!0
[
°
U
¤
(¹q ¡
w
S
)] = ¡1and upon application of the l'H^opital rule,lim
S!0
¢U = +1.
37
(11).If the transfer is conditional on the stock level (T(S);T
S
> 0) and if a fraction ® of it is embezzled
by the corrupt regulator,the problemof maximizing the government's utility has the following Hamiltonian
function:
H = ¼(p;S;Q) + (¯ ¡ 1)B(Q) + ®¯T(S) + U(S) ¡ P(S) + ¹[F(S) ¡ Q ¡
¹
H(S)]:and the ¯rst order
conditions:
dH
dQ
= 0 = ¼
Q
+(¯ ¡1)B
Q
¡¹
dH
dS
= ±¹ ¡ _¹ = ¼
S
+®¯T
S
+U
S
¡P
S
+¹(F
S
¡
¹
H
S
)
dH

=
_
S = F(S) ¡Q¡
¹
H(S):
Taking the time derivative in the ¯rst condition and substituting into the second yields:
(± ¡F
S
+
¹
H
S
)[¼
Q
+(¯ ¡1)B
Q
] ¡¼
QQ
_
Q¡(¯ ¡1)B
QQ
_
Q = ¼
S
+®¯T
S
+U
S
¡P
S
:
In the steady-state this becomes:
(± ¡F
S
¹
H
S
)[¼
Q
+(¯ ¡1)B
Q
] = ¼
S
+®¯T
S
+U
S
¡P
S
:
Similarly,solving the joint government-industry objective maximization (C
T
2
) yields:
(± ¡F
S
+
¹
H
S
)[2¼
Q
+(¯ ¡2)B
Q
] = 2¼
S
+®¯T
S
+U
S
¡P
S
:
Eliminating B
Q
between the two equations above yields:F
S
¡
¹
H
S
+
¯¼
S
+U
S
¡P
S
+®¯T
S
¯¼
Q
= ±.
To determine the e®ect of an increased marginal transfer in the form of an upward shift and/or rotation
of the transfer function,express the above condition as:
E = (F
S
¡
¹
H
S
¡±)¯¼
Q
+¯¼
S
+U
S
¡P
S
+(®¯ +1)(¿ +T
S
).In order to ¯nd the e®ect of T on the
equilibrium stock level we added the`shift parameter'¿.The e®ect of the transfer on the stock could
then be determined as:
dS
d¿
.
i
Totally di®erentiate expression E to get:
dS
d¿
= ¡
E
¿
E
S
.Since E
¿
= ®¯ > 0,we get that sign(
dS
d¿
) =
¡sign(E
S
);where E
S
= (F
SS
¡
¹
H
SS
)¯¼
Q
¡ (± ¡ F
S
+
¹
H
S
)¯¼
QS
+ ¯¼
SS
+ U
SS
+ ®¯T
SS
.From the
concavity of functions F;¼;u and T,and when S > S
00
so that P
S
S > 0,we get that expression E is
decreasing in stock level and consequently
dS
d¿
> 0:an upward shift in the transfer function increases the
conservationist e®ect of conditional aid.This formally shows the ¯rst part of Proposition 4.
We have:S
¿
= ¡
®¯
(F
SS
¡
¹
H
SS
)¯¼
Q
+(F
S
¡
¹
H
S
¡±)¯¼
QS
+¯¼
SS
+U
SS
¡P
SS
+®¯T
SS
and we can determine the mar-
ginal impacts of the corruption parameters on this conservationist e®ect.Thus,
S
¿®
= ¡
¯[(F
SS
¡
¹
H
SS
)¯¼
Q
+(F
S
¡
¹
H
S
¡±)¯¼
QS
+¯¼
SS
+U
SS
¡P
SS
+®¯T
SS
]¡®¯
2
T
SS
[(F
SS
¡
¹
H
SS
)¯¼
Q
+(F
S
¡
¹
H
S
¡±)¯¼
QS
+¯¼
SS
+U
SS
¡P
SS
+®¯T
SS
]
2
is positive,since the numerator
is positive whenever ¯[(F
SS
¡
¹
H
SS
)¯¼
Q
+(F
S
¡
¹
H
S
¡±)¯¼
QS
+¯¼
SS
+U
SS
¡P
SS
] < 0 which holds under
our assumptions on ¯ and the primitive functions.
Moreover,
i
See e.g.Skonhoft and Solstad (1998) p.30,who use a similar approach to determine comparative statics with respect to a function.
38
S
¿¯
= ¡
®[(F
SS
¡
¹
H
SS
)¯¼
Q
+(F
S
¡
¹
H
S
¡±)¯¼
QS
+¯¼
SS
+U
SS
¡P
SS
+®¯T
SS
]¡[®¯[(F
SS
¡
¹
H
SS

Q
+(F
S
¡
¹
H
S
¡±)¼
QS

SS
+®T
SS
]
[(F
SS
¡
¹
H
SS
)¯¼
Q
+(F
S
¡
¹
H
S
¡±)¯¼
QS
+¯¼
SS
+U
SS
¡P
SS
+®¯T
SS
]
2
is also positive provided that ® <
(F
SS
¡
¹
H
SS

Q
+(F
S
¡
¹
H
S
¡±)¼
QS

SS
F
SS
¼
Q
+(F
S
¡±)¼
QS

SS
+U
SS
¡P
SS
.
If,instead,the transfer is conditional on the harvest rate Q,the Hamiltonian for problem (C
T
1
)'in the
text is:
H = ¼(p;S;Q) + (¯ ¡ 1)B(Q) + ®¯T(Q) + U(S) ¡ P(S) + ¹[F(S) ¡ Q ¡
¹
H(S):] and the ¯rst order
conditions are:
dH
dQ
= 0 = ¼
Q
+(¯ ¡1)B
Q
+®¯T
Q
= ¹
dH
dS
= ±¹ ¡ _¹ = ¼
S
+U
S
¡P
S
+¹(F
S
¡
¹
H
S
)
dH

=
_
S = F(S) ¡Q¡
H(S):
Taking the time derivative in the ¯rst condition and substituting into the second yields:(± ¡ F
S
+
¹
H
S
)[¼
Q
+(¯ ¡1)B
Q
+®¯T
Q
] ¡¼
QQ
_
Q¡(¯ ¡1)B
QQ
_
Q¡®¯T
QQ
_
Q = ¼
S
+U
S
¡P
S
.
Program (C
T
2
)'yields a similar expression in the form:(± ¡ F
S
+
¹
H
S
)[2¼
Q
+ (¯ ¡ 2)B
Q
+ ®¯T
Q
] ¡

QQ
_
Q¡(¯ ¡2)B
QQ
_
Q¡®¯T
QQ
_
Q = 2¼
S
+U
S
¡P
S
.
Imposing the steady-state conditions in both expressions above we get:
(± ¡F
S
+
¹
H
S
)[¼
Q
+(¯ ¡1)B
Q
+®¯T
Q
] = ¼
S
+U
S
¡P
S
and (± ¡F
S
+
¹
H
S
)[2¼
Q
+(¯ ¡2)B
Q
+®¯T
Q
] =

S
+U
S
¡P
S
,which upon elimination of B
Q
further yields:F
S
¡
¹
H
S
+
¯¼
S
+U
S
¡P
S
¯¼
Q
+®¯T
Q
= ±.This policy rule
implicitly gives the steady-state stock of the resource.
To compare this with the resource level in the absence of international aid,notice that the transfer is
decreasing in harvesting quota Q or T
Q
< 0 and then the marginal stock e®ect is higher with the transfer,
ceteris paribus,representing a higher equilibrium stock.
Alternatively,we can think about a simple speci¯c functional form for the transfer:T(Q) = µ ¡ ÁQ
with both coe±cients positive,so that T
Q
= ¡Á.Plugging this into the expression for MSE
T
n=c
above
yields a lower marginal stock e®ect than in the absence of the transfer.¤
APPENDIX B:
Some of the most important variables in the dataset are detailed in the following table,which also lists
their respective sources:
ii
ii
The following abbreviations have been used:UCDP-PRIO= Uppsala Con°ict Data Project-International Peace Research Institute
Oslo,W.D.I= World Development Indicators (World Bank,2006),ANS database = Adjusted Net Savings Indicators (World Bank,
2006),PRS-ICRG = International Country Risk Guide from the Political Risk Services.
39
Table 1:Variables and sources
Variable Explanation Source
intensity Intensity of con°ict in given year:1-minor,2-war UCDP-PRIO
c
onset2 Dummy turns 1 when a con°ict breaks out in given year based on UCDP-PRIO
incon°ict Dummy turns 1 when a con°ict is ongoing in given year based on UCDP-PRIO
incon°ict5 Dummy turns 1 if con°ict in the next 5 years based on UCDP-PRIO
intensitymax5 Maximum intensity of con°ict in the next 5 years based on UCDP-PRIO
peacebefore Number of years of continuous peace before given year based on UCDP-PRIO
fuel
exp Fuel exports (% of merchandise exports) W.D.I.
pop Population (de°ated) W.D.I.
gdp
cap GDP per capita (de°ated) W.D.I.
gdp
growth GDP growth (annual %) W.D.I.
gini Gini index W.D.I.
unempl
ym Unemployment,youth male (%) W.D.I.
food
exp Food exports (% of merchandise exp.) W.D.I.
fuel
exp Fuel exports (% of merchandise exp.) W.D.I.
metalore
exp Ores and metals exports (% of merchandise exp.) W.D.I.
polity2 Revised combined polity score POLITY IV project
ans
e Adjusted Net Savings (PM10 excl.) A.N.S.
ans
end Energy Depletion A.N.S.
ans
i Adjusted Net Savings (PM10 incl.) A.N.S.
ans
mid Minerals Depletion A.N.S.
ans
nfd Net Forest Depletion A.N.S.
cor Corruption PRS-ICRG
bqua Bureaucratic Quality PRS-ICRG
Table 3 in the text explores the empirical relationship between the quality of governance and the level
of environmental depletion in a country.Table 3.1 below shows the same relationships hold when the
governance quality bqua is combined with the environmental indicator anse,while the same is not true
for the control of corruption indicator cor when used as a determinant for nfd.This can be tentatively
interpreted as suggesting that corruption control has more systemic implications and is linked to higher
level decisions,while bureaucratic quality may a®ect all levels of environmental decision-making.
Arti¯cially drawn post-colonial borders are thought to favour ethnic and territorial disputes,especially
in Africa.Consequently,the database was updated with three measures of fractionalization taken from
Alesina et al.(2003),referring to ethnic,language and religious dimensions.This is the most comprehensive
source available and the three indices range from zero (perfect homogeneity) to one (perfect heterogeneity
or fractionalization) are calculated as:
frac
j
= 1 ¡
n
X
i=1
s
2
ij
40
Table 2(3.1):Depletion and governance quality (bqua and anse,cor and nfd)
Variables Dependent Variable: anse Dependent Variable: nfd
(1)OLS (2) OLS (3) FE (4) OLS (5)OLS (6)FE

Bur.Quality (lag) 1.689*** 1.499*** 2.502***
(0.277) (0.292) (0.798)
Corruption (lag) 0.029** 0.013 0.119***
(0.012) (0.012) (0.037)
Gdp/capita (lag) 0.116 -0.201 -0.015*** -0.013
(0.078) (0.426) (0.005) (0.025)
Population (ln,lag) 1.758*** -5.631 0.148** 0.404*
(0.629) (4.496) (0.060) (0.237)
Polity (lag) 0.096** -0.076 0.004* -0.010
(0.046) (0.124) (0.003) (0.007)
Inconflict (lag) -0.548 -0.154**
(1.265) (0.071)
Inconflict5 (lag) 2.758 0.101
(1.763) (0.098)
Constant 1.768 -7.705** 28.855 0.331*** -0.206 -1.597
(1.194) (3.237) (24.304) (0.127) (0.317) (1.290)
Observations 2114 2029 536 2317 2150 591
Groups 122 117 47 131 121 51

Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
Table 3(5.2):Civil con°ict onset
Variables Dependent variable: conflict onset
(1)PanelProbit (2)PanelProbit (3)PanelProbit (4)PanelProbit (5)PanelProbit (6)PanelProbit

Gdp / cap (lag) -0.014 -0.018 -0.015 -0.002 -0.002 -0.018
(0.015) (0.016) (0.017) (0.017) (0.017) (0.015)
Population (ln,lag) 0.182 0.145 0.157 0.207 0.207 0.146
(0.066)*** (0.064)** (0.063)** (0.067)*** (0.068)*** (0.064)**
Corruption (lag) -0.058 -0.100 -0.100
(0.084) (0.089) (0.089)
anse (lag) -0.003 -0.008 -0.008
(0.014) (0.011) (0.011)
Corruption*anse (lag) -0.004
(0.005)
Polity (lag) 0.010 0.012 0.011 0.011 0.010 0.012
(0.015) (0.015) (0.014) (0.014) (0.014) (0.015)
year 0.013 0.012 0.014 0.015 0.015 0.012
(0.014) (0.013) (0.012) (0.013) (0.013) (0.013)
Net forest depletion (lag) -0.149 0.078 -0.152
(0.162) (0.067) (0.161)
Corruption*nfd (lag) 0.099 0.100
(0.069) (0.068)
Bur. Quality (lag) -0.123 -0.167 -0.166
(0.110) (0.113) (0.113)
Gdp growth (lag) -0.001 -0.007
(0.017) (0.015)
Constant -28.696 -25.879 -30.786 -32.643 -32.882 -27.021
(27.293) (26.579) (24.866) (25.972) (26.245) (26.725)
Bur. Quality*anse (lag) -0.004 -0.004
(0.007) (0.007)
Bur. Quality*nfd (lag) -0.020
(0.064)
Observations 2128 2266 2266 2128 2127 2262
Groups 117 121 121 117 117 121

Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
where s
ij
is the share of group i in country j.
iii
Fractionalization is interpreted as the probability that
two randomly chosen individuals belong to di®erent ethnic,linguistic or religious group.Note that these
iii
i.e.Fractionalization equals 1 minus the respective group Her¯ndahl concentration index.See Alesina et al.(2003),p.159.
41
are time-invariant measures
iv
and therefore justify the use of the random e®ects probit model.As El-
badawi and Sambanis note,a ¯xed e®ects speci¯cation would create multicollinearity between the ¯xed
individually-speci¯c error-component and the time invariant regressor.
v
Results are presented in Table 4.
Table 4:Civil con°ict incidence and measures of fractionalization
Variables Dependent variable: inconflict5
(1)PanelProbit (2)PanelProbit (3)PanelProbit (4)PanelProbit (5)PanelProbit
Gdp / cap (lag) 0.000 -0.000 -0.000 -0.000 -0.000
(0.000) (0.000)* (0.000) (0.000) (0.000)
Population (ln,lag) 0.628 0.579 0.668 0.561 0.573
(0.142)*** (0.122)*** (0.135)*** (0.128)*** (0.130)***
Corruption (lag) -0.360 -0.409 -0.416 -0.365 -0.407
(0.087)*** (0.085)*** (0.085)*** (0.086)*** (0.085)***
nfd (lag) -0.112 -0.110 -0.141 -0.118 -0.094
(0.178) (0.158) (0.167) (0.178) (0.157)
Corruption *nfd (lag) 0.091 0.107 0.105 0.093 0.107
(0.059) (0.058)* (0.058)* (0.059) (0.058)*
Polity (lag) 0.015 0.003 0.008 0.013 0.001
(0.015) (0.014) (0.014) (0.015) (0.014)
year -0.119 -0.112 -0.116 -0.118 -0.112
(0.012)*** (0.011)*** (0.012)*** (0.012)*** (0.011)***
Ethnic fractionalization 1.732 2.064
(0.957)* (0.670)***
Language fractionalization 2.438 2.540
(0.803)*** (0.699)***
Religion fractionalization -2.083 -0.919
(0.894)** (0.842)
Constant 233.219 221.120 226.425 230.353 220.751
(24.349)*** (22.325)*** (23.131)*** (23.933)*** (22.289)***
Observations 2210 2266 2251 2225 2266
Groups 118 121 120 119 121
Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
The second column in Table 4 contains the general panel probit speci¯cation including lagged regres-
sors and a time trend,columns three to ¯ve add -in turn - measures of ethnic fractionalization,language
fractionalization and religious fractionalization,respectively,while column one includes all three dimen-
sions.First,it can be seen that the relationships of interest to us remain signi¯cant in all columns
vi
and
change little in magnitude when the degree of fragmentation in the society is controlled for.Second,the
measures of ethnic and linguistic fractionalization in a society have a positive in°uence on the likelihood
of con°ict:
vii
as hypothesized,some con°icts indeed seem to have roots in ethnic divisions.Third,reli-
gious fractionalization indicator has the opposite sign but is not statistically signi¯cant.This result is
iv
Although there is some evidence that these fractionalization measures are in principle endogenous in the long-run,due to diverse
factors such as:di®erences in fertility rates across groups,migration,mixing,de¯nitional changes and identity/a±liation shifts,they
are shown to be very stable in a time frame of up to 30 years.See Alesina et al.(2003),p.161.
v
See Elbadawi and Sambanis (2002),p.313.
vi
Though the level of signi¯cance of the interaction term drops to 5% in the ¯rst and fourth columns.
vii
This result con¯rms the ¯ndings of Elbadawi and Sambanis (2002),Hodler (2006).The latter paper's result that fractionalization
is linked to the perpetuation of resource-related violence is driven by the assumption that the number of rival groups in the society is
inversely proportional to property rights enforcement,and thus the productivity in legitimate economic activities.
42
consistent with other studies in the literature and is usually explained by the fact that religious a±liation
is`more endogenous'than the other measures,as it is relatively easy to hide or change under an intolerant
regime.
viii
Another factor that appears to play a role in some of the existing empirical studies of civil con°ict is
the previous history of violence.We added the peacebefore variable,that measures the number of years
at peace (since 1946) before the observation year.As it can be seen in Table 8 (column 4),peacebefore
is negative and strongly signi¯cant in a speci¯cation that has inconflict5 as a dependent variable,while
none of the other regressors,except for population and level of democracy,are signi¯cant any longer,
although they keep the hypothesized signs.While this is a discouraging result,showing that violence
breeds violence,it is also di±cult to interpret,since the period of peace before the observation year is
jointly determined with all the other regressors.
Table 5(5.3):Robustness checks:di®erent models
Variables Dependent variable: inconflict5
(1)PanelProbit (2)PanelProbit,pa (3)Panel gee (4)PanelLogit (5)PanelLogit, fe (6) PanelLogit, fe

Gdp growth (lag) -0.029 -0.007 -0.003 -0.042 -0.070
(2.98)* (1.80)*** (2.92)* (2.82)* (3.21)*
Population (ln,lag) 0.539 0.202 0.057 0.889 2.005 12.910
(4.23)* (3.08)* (2.93)* (4.79)* (0.90) (3.85)*
Corruption (lag) -0.444 -0.153 -0.044 -0.763 -0.607 -0.569
(5.20)* (5.27)* (5.19)* (5.41)* (3.62)* (3.27)*
Net Forest Depletion
(lag)
-0.130 0.010 0.005 -0.219 -0.825 -0.845
(0.81) (0.13) (0.22) (0.82) (2.04)** (2.14)**
Corruption*nfd 0.133 0.018 0.003 0.232 0.195 0.238
(2.23)** (0.72) (0.40) (2.39)** (1.82)*** (2.26)**
Polity (lag) 0.000 -0.007 -0.003 -0.004 0.029 0.032
(0.02) (1.27) (1.94)*** (0.19) (1.05) (1.14)
Year -0.114 -0.042 -0.012 -0.191 -0.227 -0.526
(10.17)* (10.18)* (10.41)* (10.09)* (4.46)* (5.98)*
Gdp/capita (lag) 0.001
(4.53)*
Constant 225.016 81.715 23.422 375.984
(10.07)* (10.18)* (10.56)* (9.98)*
Observations 2289 2289 2289 2289 733 735
Groups 124 124 124 124 36 36

Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
viii
e.g.Alesina et al.(2003),p.167.
43
Table 6(5.4):Robustness checks:di®erent depletion indicators
Variables Dependent variable: inconflict5
(1)PanelProbit (2)PanelProbit (3)PanelProbit (4)PanelProbit (5)PanelProbit (6)PanelProbit

Gdp/capita (lag) 0.021 -0.041 -0.055 -0.045 -0.017 -0.007
(0.026) (0.025)* (0.025)** (0.024)* (0.027) (0.027)
Population (ln,lag) 0.596 0.579 0.544 0.570 0.578 0.597
(0.131)*** (0.122)*** (0.126)*** (0.120)*** (0.127)*** (0.117)***
Corruption (lag) -0.442 -0.409 -0.353 -0.442
(0.111)*** (0.085)*** (0.079)*** (0.077)***
Bur.Quality (lag) -0.606 -0.504 -0.578
(0.135)*** (0.100)*** (0.098)***
Net Forest Dep. (lag) -0.451 -0.110 -0.036
(0.167)*** (0.158) (0.097)
Energy depl. (lag) -0.026 -0.017
(0.021) (0.019)
Mineral depl. (lag) -0.557 -0.345 -0.372
(0.109)*** (0.093)*** (0.067)***
Corruption*nfd (lag) 0.147 0.107
(0.060)** (0.058)*
Bur.Quality*nfd (lag) 0.220 0.183
(0.081)*** (0.076)**
Corruption*end (lag) 0.005 0.004
(0.007) (0.006)
Bur.Quality*end (lag) 0.006
(0.008)
Corruption*mid (lag) 0.076 0.122
(0.036)** (0.031)***
Bur.Quality*mid (lag) 0.214 0.234
(0.043)*** (0.038)***
Polity (lag) -0.014 0.003 0.006 0.000 -0.003 -0.016
(0.014) (0.014) (0.014) (0.014) (0.013) (0.014)
year -0.115 -0.112 -0.108 -0.111 -0.104 -0.104
(0.012)*** (0.011)*** (0.011)*** (0.011)*** (0.011)*** (0.011)***
Constant 227.255 221.120 212.205 218.803 204.569 203.703
(24.078)*** (22.325)*** (22.230)*** (22.838)*** (21.953)*** (22.496)***
Observations 2266 2266 2296 2296 2266 2296
Groups 121 121 122 122 121 122

Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
Table 7(5.5):Robustness checks:additional controls 1
Variables Dependent variable: inconflict5
(1)PanelProbit (2)PanelProbit (3)PanelProbit (4)PanelProbit (5)PanelProbit (6)PanelProbit

Gdp /capita (lag) -0.000 -0.000 -0.000 0.000 0.000 0.000
(0.18) (1.02) (2.18)** (1.97)** (2.24)** (1.98)**
Population (ln,lag) 0.726 0.635 0.612 0.938 0.964 0.829
(3.19)* (3.03)* (4.71)* (4.80)* (4.70)* (3.78)*
Corruption (lag) -0.180 -0.594 -0.403 -0.997 -1.017 -1.083
(0.75) (1.63) (5.18)* (4.17)* (3.82)* (4.06)*
Anse (lag) -0.015 -0.106 0.035 -0.002 -0.002 -0.022
(0.38) (1.73)*** (2.10)** (0.05) (0.04) (0.37)
Corruption*anse (lag) -0.010 0.022 -0.010 -0.011 -0.009 -0.005
(0.76) (1.03) (2.29)** (0.80) (0.50) (0.30)
Polity (lag) 0.054 0.043 -0.001 -0.052 -0.067 -0.085
(1.38) (0.84) (0.04) (1.48) (1.74)*** (2.08)**
year -0.098 -0.200 -0.103 -0.156 -0.166 -0.171
(2.83)* (2.17)** (7.83)* (4.92)* (4.96)* (4.99)*
Literacy youth male -0.020
(1.14)
Perc. Low Income -0.120
(1.07)
Rural pop density 0.000
(0.53)
Unempl. youth male 0.040 0.043 0.040
(1.57) (1.67)*** (1.53)
Food exports (%) -0.021
(1.36)
Fuel exports (%) 0.013
(0.74)
Constant 193.185 397.271 202.634 306.221 325.364 338.165
(2.80)* (2.16)** (7.75)* (4.85)* (4.89)* (4.93)*
Observations 140 104 1892 871 842 844
Groups 91 103 116 82 80 80

Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
44
-.02
0
.02
.04
Interaction Effect (percentage points)
0
.2
.4
.6
.8
1
Predicted Probability that y = 1
Correct interaction effect Incorrect marginal effect
Interaction Effects after Probit
-5
0
5
10
z-statistic
0
.2
.4
.6
.8
1
Predicted Probability that y = 1
z-statistics of Interaction Effects after Probit
Table 8(5.6):Interaction e®ect bqua,nfd (Based on probit regression in Table5 col.2 in text)
Variables Dependent variable: inconflict5
(1)PanelProbit (2)PanelProbit (3)PanelProbit (4)PanelProbit (5)PanelProbit (6)PanelProbit

Gdp / capita (lag) -0.000 0.000 -0.000 0.000 -0.000 -0.000
(0.44) (1.60) (0.30) (0.77) (1.71)*** (2.02)**
Population (ln,lag) 0.928 0.570 0.623 -0.171 0.560 0.767
(3.52)* (3.58)* (4.45)* (0.83) (4.32)* (4.84)*
Corruption (lag) -1.130 -0.859 -0.584 -0.221 -0.373 -0.493
(3.85)* (6.16)* (5.81)* (1.18) (4.89)* (4.98)*
anse -0.106 -0.007 0.024 0.034 0.035 0.012
(2.03)** (0.22) (1.11) (0.97) (2.20)** (0.61)
Corruption*anse (lag) 0.022 -0.005 -0.008 -0.014 -0.011 -0.008
(1.24) (0.61) (1.41) (0.97) (2.47)** (1.40)
Polity (lag) 0.049 -0.031 -0.005 0.036 -0.002 0.009
(1.23) (1.17) (0.29) (1.05) (0.15) (0.49)
Year -0.135 -0.164 -0.133 0.212 -0.105 -0.135
(3.56)* (7.57)* (8.35)* (5.00)* (8.59)* (7.57)*
Gini coefficient 0.024
(0.81)
Total unemployment -0.017
(0.59)
Metal ore exports 0.016
(1.20)
Arms imports 0.000
(0.18)
Exports/imports 0.000
(0.32)
Energy depletion -0.112
(1.53)
Peacebefore -0.063
(2.44)**
Constant 266.227 325.317 262.134 -418.605 207.287 266.592
(3.52)* (7.52)* (8.28)* (4.95)* (8.51)* (7.48)*
Observations 294 1155 1666 530 2014 1579
Groups 102 97 116 46 117 112

Note: Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
Table 9(5.7):Robustness checks:additional controls 2
References
[1]
Acemoglu,D.,Johnson,S.,and Robinson,J.A.The colonial origins of comparative
development:An empirical investigation.The American Economic Review,5,1369-1401.
[2]
Acemoglu,D.,and Robinson,J.A.Economic Origins of Dictatorship and Democracy.
Cambridge University Press:New York,2005.
[3]
Alesina,A.,Devleeschauwer,A.,Easterly,W.,Kurlat,S.,and Wacziarg,R.Frac-
tionalization.Journal of Economic Growth 8 (2003),155-194.
[4]
Aslaksen,S.,and Torvik,R.A theory of civil con°ict and democracy in rentier states.
Working paper series,Department of Economics,Norwegian University of Science and Tech-
nology,May 2005.
[5]
Azam,J.-P.The redistributive state and con°icts in africa.Journal of Peace Research 38,4
(2001),429.
[6]
Azam,J.-P.,and Mesnard,A.Civil war and the social contract.Public Choice 115,34
(2003),455-75.
[7]
Baland,J.-M.,and Francois,P.Rent-seeking and resource booms.Journal of Development
Economics 61,2 (April 2000),527-542.
[8]
Baland,J.-M.,and Platteau,J.-P.Halting degradation of natural resources:is there a
role for rural communities?New York:Food and Agriculture Organization of the United
Nations;Oxford,England:Clarendon Press,1996.
[9]
Ballentine,K.,and Sherman,J.,Eds.The Political Economy of Armed Con°ict:Beyond
Greed and Grievance.Lynne Rienner Publishers,London,2003.
[10]
Bannon,I.,and Collier,P.,Eds.Natural Resources and Violent Con°ict:Options and
Actions.The World Bank:Washington,D.C.,2003.
[11]
Barbier,E.,Damania,R.,and Leonard,D.Corruption,trade and resource conversion.
Journal of Environmental Economics and Management 50,2 (2005),276-299.
[12]
Barbier,E.B.,and Rauscher,M.Trade,tropical deforestation and policy interventions.
Environmental and Resource Economics 4,1 (1994),75-90.
[13]
Barbier,E.B.,and Schulz,C.-E.Wildlife,biodiversity and trade.Environment and De-
velopment Economics 2,02 (March 2001),145-172.
[14]
Bardhan,P.Scarcity,Con°icts,and Cooperation.MIT Press,Cambridge,MA,2005.
[15]
Bergemann,D.,and Valimaki,J.Dynamic common agency.Discussion papers,Northwest-
ern University,Center for Mathematical Studies in Economics and Management Science,Nov.
1999.
[16]
Bernheim,B.D.,and Whinston,M.D.Menu auctions,resource allocation,and economic
in°uence.The Quarterly Journal of Economics 101,1 (February 1986),1{31.
[17]
Bockstette,V.,Chanda,A.,and Putterman,L.States and markets:The advantage of
an early start.Journal of Economic Growth 7,4 (December 2002),347{69.
[18]
Bollen,K.A.,Guilkey,D.K.,and Mroz,T.A.Binary outcomes and endogenous ex-
planatory variables:Tests and solutions with an application to the demand for contraceptive
use in tunisia.Demography 32,1 (Feb 1995),111-131.
[19]
Bolt,K.,Matete,M.,and Clemens,M.Manual for Calculating Adjusted Net Savings.
Environment Department,World Bank,September 2002.
[20]
Borrmann,A.,Busse,M.,and Neuhaus,S.Institutional quality and the gains from trade.
Kyklos 59,3 (August 2006),1-26.
[21]
Brunnschweiler,C.N.,and Bulte,E.H.The resource curse revisited and revised:A
tale of paradoxes and red herrings.Journal of Environmental Economics and Management
forthcoming (2007).
[22]
Caselli,F.,and Coleman,W.J.On the theory of ethnic con°ict.July 2002.
[23]
Clark,C.W.Pro¯t maximization and the extinction of animal species.Journal of Political
Economy 81,4 (July-Aug.1973),950-61.
[24]
Clark,C.W.,and Munro,G.R.The economics of ¯shing and modern capital theory:A
simpli¯ed approach.Journal of Environmental Economics and Management 2,2 (1975/12),
92-106.
[25]
Collier,P.,and Hoeffler,A.On economic causes of civil war.Oxford Economic Papers
50,4 (1998),563-73.
[26]
Collier,P.,and Hoeffler,A.Greed and grievance in civil war.Development and comp
systems,Economics Working Paper Archive at WUSTL,Sept.2004.
i
ii
[27]
Collier,P.,and Hoeffler,A.Resource Rents,Governance and Con°ict.Journal of Con-
°ict Resolution 49,4 (2005),625-633.
[28]
Copeland,B.R.Strategic enhancement and destruction of ¯sheries and the environment in
the presence of international externalities.Journal of Environmental Economics and Man-
agement 19,3 (1990),213-226.
[29]
Cramer,C.Homo economicus goes to war:Methodological individualism,rational choice
and the political economy of war.World Development 30 (November 2002),1845-1864.
[30]
DalBo,P.,and DalBo,E.Workers,warriors and criminals:Social con°ict in general
equilibrium.Working papers,Brown University,Department of Economics,2005.
[31]
Damania,R.Trade and the political economy of renewable resource management.Adelaide
University Discussion Paper 0046,2000.
[32]
Damania,R.,and Barbier,E.Lobbying,trade and renewable resource harvesting.CIES
Discussion Paper 0114,2001.
[33]
Damania,R.,and Gupta,A.The political economy of
resources and development in india.Adelaide University,
http://www.economics.adelaide.edu.au/workshops/doc/
guptaresources6.pdf.
[34]
Diehl,P.F.,and Gleditsch,N.P.,Eds.Environmental Con°ict.Westview Press,Boulder,
CO,2001.
[35]
Elbadawi,I.,and Sambanis,N.How much war will we see?explaining the prevalence of
civil war.Journal of Con°ict Resolution,46 (2002).
[36]
Eliste,P.,and Fredriksson,P.G.Environmental regulations,transfers,and trade:Theory
and evidence.Journal of Environmental Economics and Management 43,2 (March 2002),
234-250.
[37]
Fearon,J.D.Primary commodity exports and civil war.Journal of Con°ict Resolution 49,
4 (2005),483-507.
[38]
Fearon,J.D.,and Laitin,D.D.Ethnicity,insurgency,and civil war.American Political
Science Review 97 (February 2003),75-90.
[39]
Fjelde,H.Natural resources,quality of government and civil con-
°ict.Department of Peace and Con°ict Research,Uppsala University,
http://www.pcr.uu.se/publications/other
pub/isapaper
fjelde.pdf.
[40]
Fredriksson,P.G.,and Svensson,J.Political instability,corruption and policy formation:
the case of environmental policy.Journal of Public Economics 87,78 (August 2003),1383-
1405.
[41]
Garfinkel,M.R.,and Skaperdas,S.Economics of con°ict:An overview.Working papers,
University of California-Irvine,Department of Economics,2006.
[42]
Gershenson,D.,and Grossman,H.I.Civil con°ict:Ended or never ending?Journal of
Con°ict Resolution 44,6 (December 2000),807-821.
[43]
Gleditsch,N.P.Armed con°ict and the environment:A critique of the literature.Journal
of Peace Research,35 (1998),381-400.
[44]
Gleditsch,N.P.,Wallensteen,P.,Eriksson,M.,Sollenberg,M.,and Strand,H.
Armed con°ict 1946-2001:A new dataset.Journal of Peace Research 39,5 (2002),615-637.
[45]
Gonzalez,F.M.E®ective property rights,con°ict and growth.Journal of Economic Theory
(in press).
[46]
Grossman,G.M.,and Helpman,E.Protection for sale.American Economic Review 84,4
(1994),833-50.
[47]
Grossman,H.I.A general equilibrium model of insurrections.American Economic Review
81,4 (1991),912-21.
[48]
Grossman,H.I.Choosing between peace and war.Nber working papers 10180,National
Bureau of Economic Research,Dec.2003.
[49]
Hamilton,K.,and Clemens,M.Genuine savings rates in developing countries.World Bank
Economic Review 13,2 (May 1999),333-56.
[50]
Hegre,H.Disentangling democracy and development as determinants of armed con°ict.The
World Bank (2003).
[51]
Hegre,H.,and Sambanis,N.Sensitivity analysis of empirical results on civil war onset.
Journal of Con°ict Resolution 50,4 (2006),508-535.
[52]
Hodler,R.The curse of natural resources in fractionalized countries.European Economic
Review 50,6 (August 2006),1367-1386.
[53]
Homer-Dixon,T.F.Environment,Scarcity,and Violence.Princeton University Press,
Princeton N.J.,1999.
[54]
Hotte,L.,Long,N.V.,and Tian,H.International trade with endogenous enforcement of
property rights.Journal of Development Economics 62,1 (June 2000),25-54.
[55]
Humphreys,M.Natural Resources,Con°ict,and Con°ict Resolution:Uncovering the Mech-
anisms.Journal of Con°ict Resolution 49,4 (2005),508-537.
iii
[56]
Kaufmann,D.,Kraay,A.,and Mastruzzi,M.Governance matters iv:Governance indi-
cators for 1996-2004.World Bank Policy Research Working Paper 3630,2005.
[57]
Klare,M.T.Resource Wars:The New Landscape of Global Con°ict.Metropolitan Books:
New York,2001.
[58]
La Porta,R.,Lopez-de Silanes,F.,Shleifer,A.,and Vishny,R.Law and ¯nance.
Journal of Political Economy 106,6 (December 1998),1113-1155.
[59]
La Porta,R.,Lopez-de Silanes,F.,Shleifer,A.,and Vishny,R.The quality of govern-
ment.Journal of Law,Economics and Organization 15,1 (April 1999),222-79.
[60]
Le Billon,P.The political economy of war:What relief agencies need to know.Humani-
tarian Practice Network paper 33,Overseas Development Institute,London,2000.
[61]
Le Billon,P.Buying peace or fuelling war:the role of corruption in armed con°icts.Journal
of International Development 15,4 (2003),413-426.
[62]
Mauro,P.Corruption and growth.The Quarterly Journal of Economics 110,3 (August
1995),681-712.
[63]
Mauro,P.Corruption and the composition of government expenditure.Journal of Public
Economics 69,2 (1998/6/1),263-279.
[64]
Maxwell,J.W.,and Reuveny,R.Resource scarcity and con°ict in developing countries.
Journal of Peace Research 37,3,301-322.
[65]
Miguel,E.,Shanker,S.,and Sergenti,E.Economic shocks and civil con°ict:An instru-
mental variables approach.Journal of Political Economy 112,4 (August 2004),725-753.
[66]
Neary,H.M.A comparison of rent-seeking models and economic models of con°ict.Public
Choice 93,3-4 (1997),373-88.
[67]
Noailly,J.,Withagen,C.A.,and van den Bergh,J.C.Spatial evolution of social norms
in a common-pool resource game.Tech.rep.,Fondazione Eni Enrico Mattei,May 2005.
[68]
Olken,B.A.Corruption perceptions vs.corruption reality.NBER Working Papers 12428,
National Bureau of Economic Research,Inc,2006.
[69]
Olsson,O.Con°ict diamonds.Journal of Development Economics 82,2 (March 2007),
267-286.
[70]
Ramon,L.,and Mitra,S.Corruption,pollution,and the kuznets environment curve.Jour-
nal of Environmental Economics and Management 40 (2000),137-150.
[71]
Renner,M.Fighting For Survival:Environmental Decline,Social Con°ict,and the New
Age of Insecurity.Worldwatch Institute,Washington D.C.,1996.
[72]
Renner,M.The anatomy of resource wars.Worldwatch Institute,162 (2002).
[73]
Reuveny,R.,and Maxwell,J.W.Con°ict and Renewable Resources.Journal of Con°ict
Resolution 45,6 (2001),719-742.
[74]
Rivers,D.,and Vuong,Q.H.Limited information estimators and exogeneity tests for
simultaneous probit models.Journal of Econometrics 39,3 (November 1988),347-366.
[75]
Rodrik,D.,Subramanian,A.,and Trebbi,F.Institutions rule:The primacy of institutions
over geography and integration in economic development.Journal of Economic Growth 9,2
(2004),131-165.
[76]
Ron,J.Paradigm in distress?primary commodities and civil war.Journal of Con°ict Res-
olution 49,4 (2005),443-450.
[77]
Ross,M.L.The political economy of the resource curse.World Politics 51,2 (1999),297-322.
[78]
Ross,M.L.What do we know about natural resources and civil war?Journal of Peace
Research 41,3 (2004),337-356.
[79]
Schulz,C.-E.Trade policy and ecology.Environmental and Resource Economics 8,1 (1996),
15-38.
[80]
Skaperdas,S.An economic approach to analyzing civil wars.Tech.rep.,Department of
Economics,University of California,Irvine,May 2005.
[81]
Skonhoft,A.,and Solstad,J.T.The political economy of wildlife exploitation.Land
Economics 74,1 (feb 1998),16-31.
[82]
Solow,R.M.The economics of resources or the resources of economics.American Economic
Review 64,2 (May 1974),1{14.
[83]
The PRS Group,I.International country risk guide (table 3b),c,1984-Present.
http://www.prsgroup.com.
[84]
Weinstein,J.M.Resources and the information problem in rebel recruitment.Journal of
Con°ict Resolution 49,4 (2005),598-624.
[85]
Weitzman,M.L.Why the far-distant future should be discounted at its lowest possible rate.
Journal of Environmental Economics and Management 36,3 (November 1998),201-208.
[86]
Wilen,J.E.Renewable resource economists and policy:What di®erences have we made?
Journal of Environmental Economics and Management 39,3 (2000),306-327.
[87]
xxx.World development indicators.World Bank,Washington,DC.(2006).